Bug#864448: bowtie FTBFS on mips64el: /usr/bin/ld: .gnu.hash is incompatible with the MIPS ABI

[Dejan Latinovic]

There were the patch for this issue in earlier bowtie version:
https://sources.debian.net/src/bowtie/1.1.2-6/debian/patches/no_hash_style_both_for_mips.patch/

This patch prevents using hash style both for mips, in 1.2.0 version it is 
empty for some reason.

Bug#847750: rmysql: FTBFS on mips64el

[Dejan Latinovic]

Control: reassign -1  libmariadb-dev 2.3.1-1

Hi,
This issue is caused by garbage output form mariadb_config.
The issue was reported and the fix has been proposed:
https://bugs.debian.org/cgi-bin/bugreport.cgi?archive=no=850142

I will merge these two bugs.

Dejan

Bug#841431: apt-cacher-ng: FTBFS on mips/mipsel: undefined reference to '__atomic_fetch_add_8'

[Dejan Latinovic]

Package: apt-cacher-ng
Version: 1-1 
Severity: serious
Tags: sid, patch
Justification: FTBFS
User: debian-m...@lists.debian.org
Usertags: mips-patch


Build for mips and mipsel fails during apt-cacher-ng linking with error:
/usr/include/c++/6/bits/atomic_base.h:514: error: undefined reference to 
'__atomic_fetch_add_8'

Full build log is available here:
https://buildd.debian.org/status/fetch.php?pkg=apt-cacher-ng=mipsel=1-1=1476932907


libatomic1 provides 8-byte atomic operations for 32-bit MIPS, so linking 
apt-cacher-ng against latomic solves this issue.

The patch that includes this fix is attached.

Regards,
Dejan
--- ./debian/rules.orig	2016-10-20 14:27:34.0 +
+++ ./debian/rules	2016-10-20 14:58:43.0 +
@@ -11,11 +11,18 @@
 CFLAGS += $(CPPFLAGS)
 CXXFLAGS += $(CPPFLAGS)
 
+DEB_HOST_ARCH ?= $(shell dpkg-architecture -qDEB_HOST_ARCH)
+
+# libatomic provides 8-bytes atomic operation for 32-bit MIPS
+ifneq (,$(filter $(DEB_HOST_ARCH), mips mipsel))
+ EXTRA_LIBS="-DEXTRA_LIBS_ACNG=-latomic"
+endif
+
 %:
 	dh $@ --parallel --with systemd
 
 override_dh_auto_configure:
-	dh_auto_configure -- -DSDINSTALL=on
+	dh_auto_configure -- -DSDINSTALL=on $(EXTRA_LIBS)
 
 # needs extra build deps and internet -> ignore
 override_dh_auto_test:

Bug#841302: ngraph-gtk: FTBFS on mips*: ngraph segfaults on testing

[Dejan Latinovic]

Package: ngraph-gtk
Version: 6.06.13-5.1
Severity: serious
Tags: sid, patch
Justification: FTBFS
User: debian-m...@lists.debian.org
Usertags: mips-patch


Package ngraph-gtk FTBFS for mips* on testing with an error:

> make[6]: Entering directory '/«PKGBUILDDIR»/doc/html/ja'
> cat object.html.in > object.html
> ../../../src/ngraph -i ../objlist.nsc >> object.html
> Unable to init server: Could not connect: Connection refused
> /bin/bash: line 1: 23170 Segmentation fault  ../../../src/ngraph -i 
> ../objlist.nsc >> object.html
> Makefile:750: recipe for target 'object.html' failed
> make[6]: *** [object.html] Error 139

Full build log:
https://buildd.debian.org/status/fetch.php?pkg=ngraph-gtk=mipsel=6.06.13-5.1%2Bb2=1476151762


Back trace from gdb shows that segfault happens in  __GI__IO_wsetb, libc.so.6

Full back trace:
(gdb) bt
#0  __GI__IO_wsetb (f=0x76df0dbc <_IO_stdout_>, b=0x0, eb=0x0, a=0) at 
wgenops.c:105
#1  0x76cea24c in _IO_unbuffer_all () at genops.c:924
#2  _IO_cleanup () at genops.c:966
#3  0x76ca4eac in __run_exit_handlers (status=0, listp=0x76df03cc 
<__exit_funcs>, run_list_atexit=, run_dtors=) at 
exit.c:96
#4  0x76ca4f70 in __GI_exit (status=) at exit.c:105
#5  0x005051e8 in sysdone (obj=0x632988, inst=0x632ed0, rval=0x632ef0, argc=2, 
argv=0x646048) at osystem.c:171
#6  0x004a51c0 in delobj (obj=0x632988, delid=0) at object.c:2398
#7  0x00521e44 in cmdel (nshell=0x642d70, argc=2, argv=0x645f28) at 
shellcm.c:920
#8  0x0051b84c in cmdexec (nshell=0x642d70, cmdroot=0x6455d8, namedfunc=0) at 
shell.c:3906
#9  0x0051c610 in cmdexecute (nshell=0x642d70, cline=0x0) at shell.c:4115
#10 0x00501a64 in cmdshell (obj=0x631ab0, inst=0x642ca0, rval=0x642cc8, argc=3, 
argv=0x645638) at oshell.c:164
#11 0x004a6db0 in exeobj (obj=0x631ab0, vname=0x5ba400 "shell", id=0, argc=1, 
argv=0x7fff654c) at object.c:2794
#12 0x0040db10 in main (argc=3, argv=0x7fff6634) at main.c:1007


The reason for this behavior is using of version script during creation of 
ngraph executable.
Marking  _IO_stdin_used symbol as global solves this issue.

Similar problem was noticed in lua5.2 package:
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=816059


The patch that solves this is attached.

Regards,
Dejan

--- ngraph-gtk-6.06.13.orig/src/ngraph_export.map
+++ ngraph-gtk-6.06.13/src/ngraph_export.map
@@ -1,5 +1,6 @@
 {
   global:
 ngraph_*;
+_IO_stdin_used;
   local: *;
 };

Bug#834186: cpp-netlib: FTBFS: 94% tests passed, 1 tests failed out of 17

[Dejan Latinovic]

Control: tags -1 + patch

The same behavior is noticed on other architectures as well: alpha, arm64, 
kfreebsd-amd64, kfreebsd-i386, mips64el, s390x, sparc64.
https://buildd.debian.org/status/logs.php?pkg=cpp-netlib=0.11.2%2Bdfsg1-2%2Bb2=sid

I took a look on github repository and this test has been disabled in the 
latest cpp-netlib-0.12.0-final release.
Here are relevant commits:
https://github.com/cpp-netlib/cpp-netlib/commit/cb9a7fa4439b19a2447db1428d7c55187cbe4961
https://github.com/cpp-netlib/cpp-netlib/commit/6686ebd57442279e79e08c8900116e6fedf2aa9f


Could we do the same?

The patch that disables the test is attached.
With this patch I was able to build cpp-netlib successfully for amd64.

From cb9a7fa4439b19a2447db1428d7c55187cbe4961 Mon Sep 17 00:00:00 2001
From: Glyn Matthews 
Date: Sat, 30 Jan 2016 22:44:38 +0100
Subject: [PATCH] Mime tests don't actually pass, so I will disable them.

---
 CMakeLists.txt | 14 +++---
 1 file changed, 7 insertions(+), 7 deletions(-)

diff --git a/CMakeLists.txt b/CMakeLists.txt
index 2ce87cc..681f7a4 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -9,7 +9,7 @@ project(CPP-NETLIB)
 
 option( CPP-NETLIB_BUILD_SHARED_LIBS "Build cpp-netlib as shared libraries." OFF )
 option( CPP-NETLIB_BUILD_TESTS "Build the cpp-netlib project tests." ON)
-option( CPP-NETLIB_BUILD_EXPERIMENTS "Build the cpp-netlib project experiments." ON)
+# option( CPP-NETLIB_BUILD_EXPERIMENTS "Build the cpp-netlib project experiments." ON)
 option( CPP-NETLIB_BUILD_EXAMPLES "Build the cpp-netlib project examples." ON)
 option( CPP-NETLIB_ENABLE_HTTPS "Build cpp-netlib with support for https if OpenSSL is found." ON)
 
@@ -102,12 +102,12 @@ if (Boost_FOUND)
   add_subdirectory(deps/googletest)
   add_subdirectory(libs/network/test)
 endif (CPP-NETLIB_BUILD_TESTS)
-if (CPP-NETLIB_BUILD_EXPERIMENTS)
-  add_subdirectory(libs/network/experiment)
-endif (CPP-NETLIB_BUILD_EXPERIMENTS)
-if (NOT MSVC AND CPP-NETLIB_BUILD_TESTS)
-  add_subdirectory(libs/mime/test)
-endif(NOT MSVC AND CPP-NETLIB_BUILD_TESTS)
+# if (CPP-NETLIB_BUILD_EXPERIMENTS)
+#   add_subdirectory(libs/network/experiment)
+# endif (CPP-NETLIB_BUILD_EXPERIMENTS)
+# if (NOT MSVC AND CPP-NETLIB_BUILD_TESTS)
+#   add_subdirectory(libs/mime/test)
+# endif(NOT MSVC AND CPP-NETLIB_BUILD_TESTS)
 if (CPP-NETLIB_BUILD_EXAMPLES)
   add_subdirectory(libs/network/example)
 endif (CPP-NETLIB_BUILD_EXAMPLES)

Bug#836327: gsequencer: FTBFS on mips and powerpc: ags_functional_audio_test segmentation fault

[Dejan Latinovic]

Any updates on this issue?

I have tested newer releases (0.7.73 and 0.7.83) on local mips machine, both 
build failed with  Segmentation fault:

> PASS: ags_thread_test
> PASS: ags_audio_test
> PASS: ags_channel_test
> PASS: ags_recycling_test
> PASS: ags_audio_signal_test
> PASS: ags_recall_test
> PASS: ags_pattern_test
> PASS: ags_notation_test
> PASS: ags_automation_test
> ./test-driver: line 107: 10633 Segmentation fault  "$@" > $log_file 2>&1
> FAIL: ags_functional_audio_test

Bug#836589: fixed in icinga2 2.5.4-2

[Dejan Latinovic]

Hi,

Another solution could be to add ggc-min-expand=20 to CXXFLAGS.

The solution is tested on mips and mipsel.
A patch that contains needed changes is attached.

Regards,
Dejan





--- icinga2-2.5.4.orig/debian/rules	2016-08-26 08:11:01.0 +
+++ icinga2-2.5.4/debian/rules	2016-09-09 11:12:59.387685354 +
@@ -2,6 +2,13 @@
 
 #export DH_VERBOSE=1
 
+ 
+DEB_BUILD_ARCH ?= $(shell dpkg-architecture -qDEB_BUILD_ARCH)
+
+ifneq (,$(filter $(DEB_BUILD_ARCH),mips mipsel))
+	export DEB_CXXFLAGS_MAINT_APPEND = --param ggc-min-expand=20
+endif
+
 %:
 	dh $@ --with systemd --with bash_completion
 

Bug#830886: ffmpeg FTBFS for mips64el on configure

[Dejan Latinovic]

Package: ffmpeg
Version: 7:3.0.2-1
Severity: serious
Tags: sid, patch
Justification: FTBFS
User: debian-m...@lists.debian.org
Usertags: mips-patch


Package ffmpeg FTBFS for mips64el on configure with the following error:
> ERROR: LoadLibrary/dlopen not found for avisynth

Error message from config.log:
> /usr/bin/ld: /tmp/ffconf.09oJ6qCG.o: linking -mnan=2008 module with previous 
> -mnan=legacy modules
> /usr/bin/ld: failed to merge target specific data of file 
> /tmp/ffconf.09oJ6qCG.o
> collect2: error: ld returned 1 exit status

The reason for that is usage of -mips64r6 flag that triggers -mnan=200 flag.

Solution is to disable mips64r6 in debian rules.
Patch that includes this fix is attached.

I will rebuild locally latest ffmpeg version (7:3.1.1-1) for 
mips/mipsel/mips64el to confirm that patch fixes mips64el and does not break 
mips/mipsel.

Regards,
Dejandiff -uNr debian.orig/rules debian/rules
--- debian.orig/rules	2016-07-12 02:53:53.0 +0200
+++ debian/rules	2016-07-12 18:47:04.548881017 +0200
@@ -169,7 +169,8 @@
 	CONFIG += --disable-mipsdsp \
 		--disable-mipsdspr2 \
 		--disable-loongson3 \
-		--disable-mips32r6
+		--disable-mips32r6 \
+		--disable-mips64r6
 endif
 ifneq (,$(filter $(DEB_HOST_ARCH),mips mipsel))
 	CONFIG += --disable-mipsfpu

Bug#818485: elfutils: FTBFS on mips and mipsel

[Dejan Latinovic]

Hi,

Are there any updates on this issue?
Any concerns for including proposed patch?

Regards,
Dejan

Bug#828034: gnss-sdr: FTBFS on i386 and amd64: cannot convert 'unsigned int*' to 'uint16_t*

[Dejan Latinovic]

tags  828034 + patch


Workaround for this issue is available here:
https://github.com/gnss-sdr/gnss-sdr/issues/31

Patch that contains backported changes from mentioned workaround is attached.
With this patch applied I was able to build gnss-sdr successfully for i386, 
amd64, mips*.


Note:
For mips* it is needed to include patch available here:
https://bugs.debian.org/828040
--- gnss-sdr-0.0.7.orig/cmake/Modules/FindVolk.cmake
+++ gnss-sdr-0.0.7/cmake/Modules/FindVolk.cmake
@@ -27,7 +27,7 @@ FIND_LIBRARY(
   ${GNURADIO_INSTALL_PREFIX}/lib
 )
 
-set(VOLK_VERSION ${PC_VOLK_VERSION})
+## set(VOLK_VERSION ${PC_VOLK_VERSION})
 
 INCLUDE(FindPackageHandleStandardArgs)
 FIND_PACKAGE_HANDLE_STANDARD_ARGS(VOLK DEFAULT_MSG VOLK_LIBRARIES VOLK_INCLUDE_DIRS)
--- gnss-sdr-0.0.7.orig/src/algorithms/acquisition/gnuradio_blocks/CMakeLists.txt
+++ gnss-sdr-0.0.7/src/algorithms/acquisition/gnuradio_blocks/CMakeLists.txt
@@ -46,6 +46,30 @@ include_directories(
  ${VOLK_GNSSSDR_INCLUDE_DIRS}
 )
 
+## Workaround for issues with the change of VOLK API
+## See https://github.com/gnuradio/volk/issues/84
+## and https://github.com/gnss-sdr/gnss-sdr/issues/31
+
+if(NOT VOLK_VERSION)
+set(VOLK_VERSION "1.2.2")
+endif(NOT VOLK_VERSION)
+
+if(VOLK_VERSION VERSION_EQUAL 1.2.2)
+find_file(VOLK_GREATER_THAN_122
+  NAMES volk/volk_32f_index_max_32u.h
+  HINTS ${CMAKE_INSTALL_PREFIX}/include
+  PATHS /usr/local/include
+/usr/include
+${CMAKE_INSTALL_PREFIX}/include
+  )
+ if(VOLK_GREATER_THAN_122)
+ set(VOLK_VERSION "1.2.3+")
+ endif(VOLK_GREATER_THAN_122)
+endif(VOLK_VERSION VERSION_EQUAL 1.2.2)
+
+if(VOLK_VERSION VERSION_GREATER 1.2.2)
+   add_definitions(-DVOLK_GT_122=1)
+endif(VOLK_VERSION VERSION_GREATER 1.2.2)
 
 if(OPENCL_FOUND)
 include_directories( ${OPENCL_INCLUDE_DIRS} )
--- gnss-sdr-0.0.7.orig/src/algorithms/acquisition/gnuradio_blocks/galileo_e5a_noncoherent_iq_acquisition_caf_cc.cc
+++ gnss-sdr-0.0.7/src/algorithms/acquisition/gnuradio_blocks/galileo_e5a_noncoherent_iq_acquisition_caf_cc.cc
@@ -424,11 +424,22 @@ int galileo_e5a_noncoherentIQ_acquisitio
 
 // initialize acquisition algorithm
 int doppler;
+
+#if VOLK_GT_122
+ uint16_t indext = 0;
+ uint16_t indext_IA = 0;
+ uint16_t indext_IB = 0;
+ uint16_t indext_QA = 0;
+ uint16_t indext_QB = 0;
+#else
 unsigned int indext = 0;
 unsigned int indext_IA = 0;
 unsigned int indext_IB = 0;
 unsigned int indext_QA = 0;
 unsigned int indext_QB = 0;
+#endif
+
+
 float magt = 0.0;
 float magt_IA = 0.0;
 float magt_IB = 0.0;
--- gnss-sdr-0.0.7.orig/src/algorithms/acquisition/gnuradio_blocks/galileo_pcps_8ms_acquisition_cc.cc
+++ gnss-sdr-0.0.7/src/algorithms/acquisition/gnuradio_blocks/galileo_pcps_8ms_acquisition_cc.cc
@@ -242,9 +242,15 @@ int galileo_pcps_8ms_acquisition_cc::gen
 {
 // initialize acquisition algorithm
 int doppler;
+#if VOLK_GT_122
+ uint16_t indext = 0;
+ uint16_t indext_A = 0;
+ uint16_t indext_B = 0;
+#else
 unsigned int indext = 0;
 unsigned int indext_A = 0;
 unsigned int indext_B = 0;
+#endif
 float magt = 0.0;
 float magt_A = 0.0;
 float magt_B = 0.0;
--- gnss-sdr-0.0.7.orig/src/algorithms/acquisition/gnuradio_blocks/pcps_acquisition_cc.cc
+++ gnss-sdr-0.0.7/src/algorithms/acquisition/gnuradio_blocks/pcps_acquisition_cc.cc
@@ -284,7 +284,11 @@ int pcps_acquisition_cc::general_work(in
 {
 // initialize acquisition algorithm
 int doppler;
+#if VOLK_GT_122
+ uint16_t indext = 0;
+#else
 unsigned int indext = 0;
+#endif
 float magt = 0.0;
 const gr_complex *in = (const gr_complex *)input_items[0]; //Get the input samples pointer
 
--- gnss-sdr-0.0.7.orig/src/algorithms/acquisition/gnuradio_blocks/pcps_acquisition_fine_doppler_cc.cc
+++ gnss-sdr-0.0.7/src/algorithms/acquisition/gnuradio_blocks/pcps_acquisition_fine_doppler_cc.cc
@@ -222,8 +222,13 @@ double pcps_acquisition_fine_doppler_cc:
 float magt = 0.0;
 float fft_normalization_factor;
 int index_doppler = 0;
+#if VOLK_GT_122
+ uint16_t tmp_intex_t;
+ uint16_t index_time = 0;
+#else
 unsigned int tmp_intex_t;
 unsigned int index_time = 0;
+#endif
 
 for (int i=0;iget_outbuf(), fft_size_extended);
 
+#if VOLK_GT_122
+ uint16_t tmp_index_freq = 0;
+#else
 unsigned int tmp_index_freq = 0;
+#endif
 volk_32f_index_max_16u(_index_freq, p_tmp_vector, fft_size_extended);
 
 //case 

Bug#790924: pandas: FTBFS: copy() got an unexpected keyword argument 'font'

[Dejan Latinovic]

Hi Yaroslav,

did you test the solution  mentioned here:
https://github.com/pydata/pandas/issues/10125 ?

Regards,
Dejan

Bug#798590: libv8-3.14: FTBFS on mipsel: Build killed with signal TERM after 150 minutes of inactivity

[Dejan Latinovic]

Hi,


I noticed that last 4 times build failed on loongson boards
with same TERM signal error message.
https://buildd.debian.org/status/logs.php?pkg=libv8-3.14=mipsel=sid

Could you please try to build libv8-3.14 on mayer 
or other non loongson board if there are any?
Maybe we should blacklist libv8-3.14 on loongson boards?

I had tried to build libv8-3,14 locally on cavium board and package was built 
successfully.


Best Regards,
Dejan

Bug#796731: freecontact: FTBFS on mips (test suite failure)

[Dejan Latinovic]

Hi,

this is obviously a consequence of mips buildd's limitations.
In past all three successfully builds had happened on ball.
I also was able to built freecontact successfully for mips on machine with FPU, 
localy.
On machine without FPU, if icme_timeout is significantly increased
tests executes successfully but need a lot of time
so the TERM signal breaks the build.

I was not able to find ball specification, so I am not sure if
it have FPU.
Are there any mips buildd's that have FPU?
If yes, maybe we could blacklist freecontact on other machines?

If no, should we disable tests for mips?
I could provide patches and test it if needed?

Regards,
Dejan

Bug#796731: freecontact: FTBFS on mips (test suite failure)

[Dejan Latinovic]

Hi,

I had tried to build freecontact locally on two different cavium boards.
On both board build failed with TERM signal error on testing.

It seems that this is the command that  causes TERM signal:
 /build/freecontact-hkgMtc/freecontact-1.0.21/src/.libs/lt-freecontact 
 --parprof=psicov --density=0 -f ./demo_1000.aln --threads 0 --debug


The results of running this command manually:


 time /build/freecontact-hkgMtc/freecontact-1.0.21/src/.libs/lt-freecontact 
 --parprof=psicov --density=0 -f ./demo_1000.aln --threads 0 --debug

 Effective parameters: --apply-gapth=0 --clustpc=62 --cov20=0 
 --estimate-ivcov=1 --gapth=0.9 --mincontsep=5 --pseudocnt=1 
 --pscount-weight=0 --rho=-1 --density=0 [shrink_lambda=0.1] 
 --output-format=pfrmat_rr
will use 2 OMP threads
 total weight (variation) of alignment = 561.595172234625
 
 seq weight loop for 1000 seqs took 0.126313 secs
 will use 2 OMP threads
 calculated column aa frequencies, gap cols = 4
 calculated pair frequency table in 9.83136 secs
 collected apc_mean[MI] = 0.165528914074767
 aa freq sum (cell) = 0.93945475299, pairfreq sum (cell) = 
 0.88196197513
 formed covariance matrix (126/126,4)
 cov matrix Cholesky... not positive definite, dmean dev: 0.0457782
 cov matrix Cholesky... not positive definite, dmean dev: 0.0412004
 cov matrix Cholesky... not positive definite, dmean dev: 0.0370804
 cov matrix Cholesky... not positive definite, dmean dev: 0.0333723
 cov matrix Cholesky... not positive definite, dmean dev: 0.0300351
 cov matrix Cholesky... not positive definite, dmean dev: 0.0270316
 cov matrix Cholesky... not positive definite, dmean dev: 0.0243284
 cov matrix Cholesky... not positive definite, dmean dev: 0.0218956
 cov matrix Cholesky... not positive definite, dmean dev: 0.019706
 cov matrix Cholesky... not positive definite, dmean dev: 0.0177354
 cov matrix Cholesky... not positive definite, dmean dev: 0.0159619
 cov matrix Cholesky... not positive definite, dmean dev: 0.0143657
 cov matrix Cholesky... not positive definite, dmean dev: 0.0129291
 cov matrix Cholesky...
 shrank covariance matrix in 15300.1 secs
 will try rho 0.00178064, exact solution
  iteration =   1Error: inverse covariance matrix estimation exceeded time 
 limit (1800 sec).
 
 real  290m35.185s
 user  37m40.530s
 sys   253m0.660s


Regards,
Dejan

Bug#796069: protobuf: FTBFS on mips{,el} (test suite failure)

[Dejan Latinovic]

It seems that this test is problematic:

 LD_LIBRARY_PATH=./src/.libs/:./gmock/gtest/lib/.libs/:./gmock/lib/.libs:./gtest/lib/.libs/
./src/.libs/protobuf-test --gtest_filter=ExtensionSetTest.RepeatedFields*
 Running main() from gtest_main.cc
 Note: Google Test filter = ExtensionSetTest.RepeatedFields*
 [==] Running 1 test from 1 test case.
 [--] Global test environment set-up.
 [--] 1 test from ExtensionSetTest
 [ RUN  ] ExtensionSetTest.RepeatedFields
 Segmentation fault



If exclude this specific test, 26 other tests failed.


[--] Global test environment tear-down
[==] 987 tests from 117 test cases ran. (4385 ms total)
[  PASSED  ] 961 tests.
[  FAILED  ] 26 tests, listed below:
[  FAILED  ] MessageTest.ParseFromFileDescriptor
[  FAILED  ] MessageTest.ParsePackedFromFileDescriptor
[  FAILED  ] TextFormatTest.Basic
[  FAILED  ] TextFormatTest.ParseBasic
[  FAILED  ] TextFormatExtensionsTest.Extensions
[  FAILED  ] TextFormatExtensionsTest.ParseExtensions
[  FAILED  ] IoTest.CompressionOptions
[  FAILED  ] CommandLineInterfaceTest.BasicPlugin
[  FAILED  ] CommandLineInterfaceTest.GeneratorAndPlugin
[  FAILED  ] CommandLineInterfaceTest.MultipleInputs
[  FAILED  ] CommandLineInterfaceTest.MultipleInputsWithImport
[  FAILED  ] CommandLineInterfaceTest.CreateDirectory
[  FAILED  ] CommandLineInterfaceTest.GeneratorParameters
[  FAILED  ] CommandLineInterfaceTest.Insert
[  FAILED  ] CommandLineInterfaceTest.PluginOutputWriteError
[  FAILED  ] CommandLineInterfaceTest.PluginOutputDirectoryNotFoundError
[  FAILED  ] CommandLineInterfaceTest.GeneratorPluginError
[  FAILED  ] CommandLineInterfaceTest.GeneratorPluginFail
[  FAILED  ] CommandLineInterfaceTest.GeneratorPluginCrash
[  FAILED  ] CommandLineInterfaceTest.PluginReceivesSourceCodeInfo
[  FAILED  ] EncodeDecodeTest.Encode
[  FAILED  ] EncodeDecodeTest.Decode
[  FAILED  ] EncodeDecodeTest.Partial
[  FAILED  ] EncodeDecodeTest.UnknownType
[  FAILED  ] BootstrapTest.GeneratedDescriptorMatches
[  FAILED  ] GeneratedDescriptorTest.IdenticalDescriptors

26 FAILED TESTS

These results are form broadcom xlp.

I will test it on cavium board as well.

Regards,
Dejan

Bug#787862: pdf-presenter-console: Fails to build on mipsel

[Dejan Latinovic]

Hi,
I had tested it on a few local MIPS boards,
it builds successfully for both mips and mipsel.

I will ask for give back.

Regards,
Dejan

--
To UNSUBSCRIBE, email to debian-bugs-rc-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org

Bug#779634: pymad: FTBFS - No 'Setup' file. Perhaps you need to run the configure script.

[Dejan Latinovic]

Hi,

the problem appears during clear phase, not on configure.
Using BEFORE_CLEAN  instead BEFORE_CONFIGURE solves
this issue for me.
The patch that contains needed changes is attached.

I had tested it on mips, mipsel and amd64.


Could you please consider including this changes?diff -uNr pymad-0.8.orig/debian/rules pymad-0.8/debian/rules
--- pymad-0.8.orig/debian/rules	2015-03-14 11:15:27.0 +
+++ pymad-0.8/debian/rules	2015-05-18 18:00:17.0 +
@@ -5,7 +5,7 @@
 # Uncomment this to turn on verbose mode.
 #export DH_VERBOSE=1
 export PYBUILD_NAME=pymad
-export PYBUILD_BEFORE_CONFIGURE={interpreter} ./config_unix.py --prefix=/usr
+export PYBUILD_BEFORE_CLEAN={interpreter} ./config_unix.py --prefix=/usr
 
 %:
 	dh $@ --with=python2 --buildsystem=pybuild

Bug#775935: geneweb: FTBFS on various arches: Please run 'make' in directory '../src' first

[Dejan Latinovic]

Hello,

I had noticed that the gui part builds with target all
even if the main target is out.
Beside that,  in gui/Makefile OCAMLC variable is redefined and ocamlc.opt is 
used instead of
ocamlc (defined in Makefile.inc).

The patch that contains needed changes is attached,
With this patch I was able to successfully build geneweb for both mips and 
mipsel.

Could you please include this patch?

Regards,
Dejandiff -uNr geneweb-6.08dfsg.orig/Makefile geneweb-6.08dfsg/Makefile
--- geneweb-6.08dfsg.orig/Makefile	2015-03-30 13:00:12.0 +
+++ geneweb-6.08dfsg/Makefile	2015-03-30 13:01:30.0 +
@@ -18,7 +18,7 @@
 	cd gwb2ged; $(MAKE) all
 	cd setup; $(MAKE) all
 	cd gwtp; $(MAKE) all
-	cd gui; $(MAKE) all
+	cd gui; $(MAKE) out
 
 opt::
 	cd wserver; $(MAKE) opt
diff -uNr geneweb-6.08dfsg.orig/gui/Makefile geneweb-6.08dfsg/gui/Makefile
--- geneweb-6.08dfsg.orig/gui/Makefile	2015-03-30 13:00:12.0 +
+++ geneweb-6.08dfsg/gui/Makefile	2015-03-30 17:43:50.0 +
@@ -4,7 +4,6 @@
 include ../tools/Makefile.inc
 
 LABLGTK2=+lablgtk2
-OCAMLC=ocamlc.opt
 OCAMLOPT=ocamlopt.opt -annot
 GWB=..
 OBJS=../src/version.cmo ../src/iovalue.cmo ../src/buff.cmo ../src/name.cmo ../src/mutil.cmo gui.cmo

Bug#770009: Backtrace for the hang

[Dejan Latinovic]

Hi,

   On which MIPS machine did you build it ?

As I mentioned before
https://bugs.debian.org/cgi-bin/bugreport.cgi?archive=nobug=770009#45

I had tested it on Cavium board without FPU.

Unfortunately, I do not have more info.
I deleted  old imagemagic chroots,
but I just had started builds
on tree different cavium boards
and I will inform you with results.


Regards,
Dejan

--
To UNSUBSCRIBE, email to debian-bugs-rc-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org

Bug#770009: Backtrace for the hang

[Dejan Latinovic]

Hi,

did anyone tried to wait some longer time to see if command will execute?

  ./magick.sh convert
 /home/fourmond/tmp/imagemagick-6.8.9.9/debian/display-im6.svg \
-background none -define filter:blur=0.75 -filter
 Gaussian -resize $SIZE \
-gravity center -extent $SIZE  \
 
 /home/fourmond/tmp/imagemagick-6.8.9.9/debian/tmp-Q16/usr/share/icons/hicolor/$SIZE/apps/display-im6.q16.png;
 \


When I tested it earlier and
tried to build package manually in chroot created with sbuild,
package built successfully.


If this is the case, I am not sure 
how thread limitation will help to
speed up the execution time..


Could we somehow avoid this converting
while building?


Best Regards,
Dejan

--
To UNSUBSCRIBE, email to debian-bugs-rc-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org

Bug#776316: [Pkg-samba-maint] Bug#776316: samba: failed to build on mips

[Dejan Latinovic]

Hi,

I had tried to build samba locally
for mips and mipsel on four different machines,
and I was not able to reproduce this error.

Maybe we should ask for give back and see
the result.

Regards,
Dejan


--
To UNSUBSCRIBE, email to debian-bugs-rc-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org

Bug#770009: imagemagick: FTBFS on mips

[Dejan Latinovic]

Hi,

My assumption is that this action on cavium boards need a lot of time to 
execute,
and that is the reason for TERM signal.

  while read SIZE; do \
   mkdir -p 
 /«PKGBUILDDIR»/debian/tmp-Q16/usr/share/icons/hicolor/$SIZE/apps/ ;\
   cd /«PKGBUILDDIR»/debian/build-quantum-Q16; \
   ./magick.sh convert /«PKGBUILDDIR»/debian/display-im6.svg \
 -background none -define filter:blur=0.75 -filter 
 Gaussian -resize $SIZE \
 -gravity center -extent $SIZE  \
  
 /«PKGBUILDDIR»/debian/tmp-Q16/usr/share/icons/hicolor/$SIZE/apps/display-im6.q16.png;
  \
  done  /«PKGBUILDDIR»/debian/display-im6.iconssize


Starting whit 8:6.8.8.9-1 version
all imagemagick builds were done on ball (broadcom board).

https://buildd.debian.org/status/logs.php?pkg=imagemagickarch=mips


Also, for mipsel 
all builds were done on non-cavium boards (broadcom, loongson).


I will try it locally confirm this assumption.


Regards,
Dejan

--
To UNSUBSCRIBE, email to debian-bugs-rc-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org

Bug#756801: haskell-bloomfilter: FTBFS almost everywhere

[Dejan Latinovic]

Control: tags -1 + patch
Control: user -1 debian-m...@lists.debian.org
Control: usertags -1 + mips-patch



Unfortunately there is no feedback from upstream.


The problem is the condition roundedBits  0x.
On 32-bit system 0x treated as Int is -1,
due to maxBound :: Int is 214748364 (0x7fff).

If the value of roundedBits is greater than 0x7fff,
it will satisfy already exisiting condition (roundedBits = 0),
due to roundedBits is Int type.


I think that removing this additional condition (reverting commit)
would be acceptable solution.


Patch that remove this condition is attached.


With this patch I was able to build haskell-bloomfilter
on mips, mipsel, i386 and amd64.

Could you please consider including these changes?


Regards,
Dejandiff -uNr bloomfilter-2.0.0.0/Data/BloomFilter/Easy.hs haskell-bloomfilter-2.0.0.0.build/Data/BloomFilter/Easy.hs
--- bloomfilter-2.0.0.0/Data/BloomFilter/Easy.hs	2014-07-05 05:57:24.0 +
+++ haskell-bloomfilter-2.0.0.0.build/Data/BloomFilter/Easy.hs	2014-11-12 14:45:52.0 +
@@ -72,7 +72,7 @@
 minimum [((-k) * cap / log (1 - (errRate ** (1 / k))), k)
  | k - [1..100]]
 roundedBits = nextPowerOfTwo (ceiling bits)
-in if roundedBits = 0 || roundedBits  0x
+in if roundedBits = 0
then Left  capacity too large to represent
else Right (roundedBits, truncate hashes)
 

Bug#754752: cython: FTBFS on mips: timeout

[Dejan Latinovic]

Hello,

I tried to build cython 0.20.2-1 locally.

On Broadcom machine,
cython builds successfully.

On Cavium machine,
build failed with TERM signal on testing.

Also, I ran tests manually,
and all tests pass on Cavium,
but it needs a lot of time for execution.

 Ran 9153 tests in 33413.962s
 
 OK
 ALL DONE


I will test new (0.21.1-1) version,
to confirm this behaviuor.


Regards,
Dejan


--
To UNSUBSCRIBE, email to debian-bugs-rc-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org

Bug#767291: blitz++: FTBFS on mips

[Dejan Latinovic]

Hello,


Maybe it is possible to blacklist blitz++ on cavium machines,
and try to build it on broadcom board (ball, swarm).

I had tested it locally.
On cavium board,
build fail with TERM signal.
If tests are ran manually, all tests pass,
but requires a lot of time.

Actually, the array test is the one that 
causes this failure.
All other tests executes significantly faster.
(1-10 min).

Duration of array test execution on cavium board is:
304m40.451s.


On broadcom board,
package builds successfully.
Duration of array test execution on broadcom board is:
12m50.570s.


Best Regards,
Dejna




From: Andreas Tille [andr...@an3as.eu]
Sent: Thursday, October 30, 2014 8:53 AM
To: 767...@bugs.debian.org
Cc: debian-m...@lists.debian.org
Subject: Re: Bug#767291: blitz++: FTBFS on mips

Hi Ivo,

thanks for the bug report.

On Wed, Oct 29, 2014 at 10:14:00PM +0100, Ivo De Decker wrote:
 package: blitz++
 version: 1:0.10-2
 severity: serious

 It seems the latest upload of blitz++ failed on mips, but it built fine
 before. This prevents migration to testing.

I'm aware of this fact ind just contacted mips porters[1].  I had the
experience with several other packages that this is due to a not as
powerful enough build machine.  The build failure [2]

...
Running examples...
E: Caught signal ‘Terminated’: terminating immediately

could leat to the conclusion that this is the case here as well.  As I
explained in my mail to the mips porters[1] my personal solution to deal
with this would be to ask ftpmasters to remove the package for mips
architecture to enable a migration to testing.  I will wait until the
end of the weekend for any response to find a better solution before
I'll do this.

Kind regards

 Andreas.


[1] https://lists.debian.org/debian-mips/2014/10/msg00021.html
[2] https://buildd.debian.org/status/package.php?p=blitz%2B%2Bsuite=sid

https://buildd.debian.org/status/fetch.php?pkg=blitz%2B%2Barch=mipsver=1%3A0.10-2stamp=1412974511

--
http://fam-tille.de


--
To UNSUBSCRIBE, email to debian-mips-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Archive: https://lists.debian.org/20141030075301.ga17...@an3as.eu

--
To UNSUBSCRIBE, email to debian-bugs-rc-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org

Bug#766832: metview: FTBFS on mipsel: libMagWrapper.a: error adding symbols: Bad value (recompile with -fPIC)

[Dejan Latinovic]

Hello,


Last successfully built metview version
it was built against magics++ 2.18.15-7.
In magics++ 2.18.15-7 there is only static libMagWrapper
library as well.
The difference is that, in last metview version, during linking of 
bin/uPlotBatch, are used flags: -fPIE -pie.

Using this flags on mips requires that
all used object files and libraries must be compiled with fPIC flag.


I had create static library libMagWrapper
that includes object files compiled with fPIC flag
and I was able to successfully build metview.


Creating libMagWrapper shared library,
that will also require using fPIC flag during
compiling object files,
should solve this issue, as well.


Best Regards,
Dejan

--
To UNSUBSCRIBE, email to debian-bugs-rc-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org

Bug#763255: freeimage: FTBFS: transupp.c:900: undefined reference to `jpeg_core_output_dimensions'

[Dejan Latinovic]

Control: tags -1 + patch

Hi,

I had tested it on mips/mipsel,
the same error appears.

This is happening because of missing
build dependency (libjpeg8-dev).


Patch that fix this issue is attached.


Best Regards,
Dejandiff -uNr freeimage-3.15.4.orig/debian/control freeimage-3.15.4/debian/control
--- freeimage-3.15.4.orig/debian/control	2014-01-20 05:25:57.0 +
+++ freeimage-3.15.4/debian/control	2014-10-01 11:12:20.0 +
@@ -5,6 +5,7 @@
  debhelper (= 8),
  dh-autoreconf,
  libjpeg-dev,
+ libjpeg8-dev,
  liblcms2-dev,
  libmng-dev,
  libopenexr-dev,

Bug#754726: liblog4ada: FTBFS on mips/mipsel due to changed symbols

[Dejan Latinovic]

Control: tags -1 + patch


Hello,

Patch that fixes this issue is attached.

This patch contains needed changes for mips mipsel and ppc64el.
(see: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=755816)


Regards,
Dejan

liblog4ada_add_symbols.debdiff
Description: liblog4ada_add_symbols.debdiff

Bug#754619: sdformat: FTBFS on almost all archs: test suite issues

[Dejan Latinovic]

Control: tags -1 + patch


Hi,

the reason for test failures is a need to
access .sdformat in HOME directory.


Patch that fix this issue is attached.

With this patch I was able to build sdformat
for i386, amd64, mips and mipsel.

Could you please consider including these changes?


Best Regards,
Dejandiff -uNr sdformat-2.0.0.orig/debian/rules sdformat-2.0.0/debian/rules
--- sdformat-2.0.0.orig/debian/rules	2014-07-02 18:27:41.0 +
+++ sdformat-2.0.0/debian/rules	2014-09-11 17:29:45.0 +
@@ -14,7 +14,7 @@
 	dh_auto_build -- doc # Generate documentation.
 
 override_dh_auto_test:
-	LC_ALL=POSIX dh_auto_test
+	HOME=$(CURDIR) LC_ALL=POSIX dh_auto_test
 
 override_dh_strip:
 	dh_strip -a --dbg-package=libsdformat2-dbg

Bug#754757: openmsx: FTBFS on mips: timeout during g++

[Dejan Latinovic]

Hello,

The error appears while compiling HQ2xScaler.cc.
Compilation finish successfully but it needs more time
than is allowed before raising TERM signal.


I had tested it locally, and on was able to reproduce
the error with signal TERM on cavium board.

On that board, time of HQ2xScaler.cc compilation is about 194 min,
while TERM appears after 150 min.


If O0 optimization is used instead of O3, for mips,
I was able to reduce time of compilation to about 20 min.

Patch that reduce optimization to O0 for mips is attached.

Could you please consider including this patch?


The other way to successfully build openmsx for mips,
could by using some other (faster) buildd machines,
beside cavium. 


Best Regards,
Dejandiff -uNr openmsx-0.10.1.orig/build/flavour-debian.mk openmsx-0.10.1/build/flavour-debian.mk
--- openmsx-0.10.1.orig/build/flavour-debian.mk	2014-08-07 13:06:40.0 +
+++ openmsx-0.10.1/build/flavour-debian.mk	2014-08-07 13:50:07.0 +
@@ -14,7 +14,12 @@
 # Do not enable -O3 to work around GCC bug http://bugs.debian.org/647552.
 CXXFLAGS += -O2
 else
-CXXFLAGS += -O3
+	ifeq ($(DEB_HOST_ARCH_CPU),mips)
+# Use -O0 to reduce time of compilation on mips https://bugs.debian.org/754757
+CXXFLAGS += -O0
+else
+CXXFLAGS += -O3
+endif
 endif
 endif
 

Bug#754623: percona-xtradb-cluster-galera-2.x: FTBFS on many archs

[Dejan Latinovic]

Hi,

The reason for this failure is a difference
in alignment of long long for MIPS ISA and IA32.
This results that size of gu_rse structure on mips is 24,
instead of expected 20 (on ia32).

As long long is first attribute in structure,
so it is safe to use pack(4).

 struct gu_rse
 {
 long long   time;
 const void* heap_ptr;
 const void* stack_ptr;
 longpid;
 };


I assume that this fix could be used for other architectures like:
armel, armhf, powerpc, sparc.
I did not have a chance to test it on those architectures,
so my changes affects only mips and mipsel.



Solving this issue, fallowing error appears:

 gcs/src/gcs.c:1161: undefined reference to `__sync_fetch_and_add_8'


Mips platform does not have 64-bit __sync_* operations.
To avoid this behaviuor it is needed to use
corresponding __atomic_* from libatomic library.


Patch that solves both issues for mips/mipsel is attached.

Could you please consider including this patch?


Best Regards,
Dejan




diff -uNr percona-xtradb-cluster-galera-2.x-175.orig/SConstruct percona-xtradb-cluster-galera-2.x-175/SConstruct
--- percona-xtradb-cluster-galera-2.x-175.orig/SConstruct	2014-08-13 12:17:02.0 +
+++ percona-xtradb-cluster-galera-2.x-175/SConstruct	2014-08-13 17:07:21.0 +
@@ -368,7 +368,7 @@
 print 'Not using boost'
 
 # Check to see if -latomic is need for GCC atomic built-ins.
-if conf.CheckLib(library='atomic', symbol='__sync_fetch_and_add_8'):
+if conf.CheckLib(library='atomic', symbol='__sync_fetch_and_add_8') or conf.CheckLib(library='atomic', symbol='__atomic_fetch_add_8'):
 conf.env.Append(LIBS=['atomic'])
 
 # asio
diff -uNr percona-xtradb-cluster-galera-2.x-175.orig/galerautils/src/gu_atomic.h percona-xtradb-cluster-galera-2.x-175/galerautils/src/gu_atomic.h
--- percona-xtradb-cluster-galera-2.x-175.orig/galerautils/src/gu_atomic.h	2014-05-08 01:08:52.0 +
+++ percona-xtradb-cluster-galera-2.x-175/galerautils/src/gu_atomic.h	2014-08-13 17:07:46.0 +
@@ -11,6 +11,8 @@
 
 #ifdef __GNUC__
 
+#if !defined(__mips__) || defined(__mips64)
+
 #define gu_sync_fetch_and_add  __sync_fetch_and_add
 #define gu_sync_fetch_and_sub  __sync_fetch_and_sub
 #define gu_sync_fetch_and_or   __sync_fetch_and_or
@@ -26,6 +28,28 @@
 #define gu_sync_xor_and_fetch  __sync_xor_and_fetch
 #define gu_sync_nand_and_fetch __gu_sync_nand_and_fetch
 
+#else  /* __mips__ */
+
+/* Mips platform does not have 64-bit __sync_* operations.
+ * so it is needed to use corresponding __atomic_* operations from libatomic library. */
+
+#define gu_sync_fetch_and_add(value_, x)  __atomic_fetch_add(value_, x, __ATOMIC_SEQ_CST)
+#define gu_sync_fetch_and_sub(value_, x)  __atomic_fetch_sub(value_, x, __ATOMIC_SEQ_CST)  
+#define gu_sync_fetch_and_or(value_, x)   __atomic_fetch_or(value_, x, __ATOMIC_SEQ_CST) 
+#define gu_sync_fetch_and_and(value_, x)  __atomic_fetch_and(value_, x, __ATOMIC_SEQ_CST)
+#define gu_sync_fetch_and_xor(value_, x)  __atomic_fetch_xor(value_, x, __ATOMIC_SEQ_CST)
+#define gu_sync_fetch_and_nand(value_, x) __atomic_fetch_nand(value_, x, __ATOMIC_SEQ_CST)
+
+
+#define gu_sync_add_and_fetch(value_, x)  __atomic_add_fetch(value_, x, __ATOMIC_SEQ_CST)
+#define gu_sync_sub_and_fetch(value_, x)  __atomic_sub_fetch(value_, x, __ATOMIC_SEQ_CST)
+#define gu_sync_or_and_fetch(value_, x)   __atomic_or_fetch(value_, x, __ATOMIC_SEQ_CST)   
+#define gu_sync_and_and_fetch(value_, x)  __atomic_and_fetch(value_, x, __ATOMIC_SEQ_CST)
+#define gu_sync_xor_and_fetch(value_, x)  __atomic_xor_fetch(value_, x, __ATOMIC_SEQ_CST)
+#define gu_sync_nand_and_fetch(value_, x) __atomic_nand_fetch(value_, x, __ATOMIC_SEQ_CST)
+
+#endif
+
 #else /* __GNUC__ */
 #error Compiler not supported
 #endif
diff -uNr percona-xtradb-cluster-galera-2.x-175.orig/galerautils/src/gu_rand.c percona-xtradb-cluster-galera-2.x-175/galerautils/src/gu_rand.c
--- percona-xtradb-cluster-galera-2.x-175.orig/galerautils/src/gu_rand.c	2014-05-08 01:08:52.0 +
+++ percona-xtradb-cluster-galera-2.x-175/galerautils/src/gu_rand.c	2014-08-13 17:07:25.0 +
@@ -16,6 +16,16 @@
 
 /*! Structure to hold entropy data.
  *  Should be at least 20 bytes on 32-bit systems and 28 bytes on 64-bit */
+
+/*  Unlike ia32, aligment of long long for MIPS ISA is 8, and size of gu_rse is 24.
+ *  As long long atribute is first in gu_rse structure there is no harm in use pack(4) to avoid
+ *  undexpected behavior while using gu_rse structure. */
+
+#if defined(__mips__)  !defined(__mips64__)
+#pragma pack(push)
+#pragma pack(4)
+#endif
+
 struct gu_rse
 {
 long long   time;
@@ -24,6 +34,10 @@
 longpid;
 };
 
+#if defined(__mips__)  !defined(__mips64__)
+#pragma pack(pop)
+#endif
+
 typedef struct gu_rse gu_rse_t;
 
 long int

Bug#756788: libwebp: FTBFS on mips*

[Dejan Latinovic]

Adding -mips32r2 is not a good idea,
since all Debian on mips* is built for mips2.

Instead, i think that better solution would be to
define  WEBP_USE_MIPS32 only in mips32 case.
(src/dsp/dsp.h)

The patch that contains needed changes is attached.

With this patch I was able to successfully build libwebp for both mips and 
mipsel.


Could you please consider including this patch?--- libwebp-0.4.1.orig/src/dsp/dsp.h	2014-07-29 00:22:32.0 +
+++ libwebp-0.4.1/src/dsp/dsp.h	2014-08-04 12:10:23.0 +
@@ -60,7 +60,7 @@
 #define WEBP_USE_NEON
 #endif
 
-#if defined(__mips__)
+#if defined(__mips__)  defined(_MIPS_ARCH_MIPS32)
 #define WEBP_USE_MIPS32
 #endif
 

Bug#756910: libapache2-mod-auth-openidc: FTBFS on mips, powerpc, s390x (and sparc): test failure

[Dejan Latinovic]

Hello,

The reason for this failure is the fact
that p does not get info about
Additional Authentication Data length (al)
in src/jose/apr_jwe.c for big endian.

Patch that fixes this issue is attached.
This patch should enable build of libapache2-mod-auth-openidc
on big endian architectures.
I successfully built it on mips (BE).
Could you please consider including this patch?


I had already forwarded it upstream:
https://github.com/pingidentity/mod_auth_openidc/pull/21


Best Regards,
Dejandiff -uNr libapache2-mod-auth-openidc-1.5.3.orig/src/jose/apr_jwe.c libapache2-mod-auth-openidc-1.5.3/src/jose/apr_jwe.c
--- libapache2-mod-auth-openidc-1.5.3.orig/src/jose/apr_jwe.c	2014-07-20 17:11:41.0 +
+++ libapache2-mod-auth-openidc-1.5.3/src/jose/apr_jwe.c	2014-08-05 07:50:49.0 +
@@ -371,6 +371,8 @@
 		// little endian machine: reverse AAD length for big endian representation
 		for (i=0; i  sizeof(int64_t); ++i) p[sizeof(uint64_t)-1-i] = src[i];
 	}
+	else
+		memcpy(p, al, sizeof(uint64_t));
 
 //	uint64_t big_endian = htobe64(al);
 //	memcpy(p, big_endian, sizeof(int64_t));

Bug#756788: libwebp: FTBFS on mips*

[Dejan Latinovic]

Hi,
I am preparing a patch that fixes this issue.

--
To UNSUBSCRIBE, email to debian-bugs-rc-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org

Bug#751277: python-biopython: FTBFS on mips* powerpc s390x

[Dejan Latinovic]

Hello,
I took a look at test failures on mips/mipsel.

Test that fails is based on a result of dnal tool.
dnal is part of a debian package wise.

In file: Bio/Wise/__init__.py,
an error was raised because the return value from dnal
is greater than 1.

If you look at dnal.c:

   alb = make_align_dnaalign(one,two,mat,dse,-gap,-ext,-gap,-ext,dpri);
 
   printf(Score %d\n,alb-score);
 
   if( alb == NULL )
 fatal(Could not build alignment!);
 
   if( show_pretty == TRUE )
  write_pretty_seq_align(alb,one,two,12,60,stdout);
 
   if( show_alb == TRUE )
 dump_ascii_AlnBlock(alb,stdout);
 }

you can see that there is no return statement.
In this case the return value of dnal usually will be the return value 
of last used function.
This behavior is mutable.
It can not be guaranteed on all architectures,
or if different optimization flags are used.
Further, using dnal with different options, last used function
could be changed.

Adding
 return 0;
at the end of main function in dnal.c,
test_align (test_Wise.TestWise) successfully passes
on mips/mipsel.


With this fix, package python-biopython
successfully builds for mipsel.
Patch that includes this fix is attached.


Unfortunately, on mips (BE)
test test_Cluster.py failed under Python 3.4.
Any progress on this issue?


Best Regards,
Dejan

--- wise-2.4.1.orig/src/models/dnal.c	2004-11-24 16:26:11.0 +
+++ wise-2.4.1/src/models/dnal.c	2014-07-31 14:53:54.0 +
@@ -75,6 +75,8 @@
 
   if( show_alb == TRUE )
 dump_ascii_AlnBlock(alb,stdout);
+
+  return 0;
 }
 
 

Bug#741288: [Debian-med-packaging] Bug#741288: seqan: FTBFS on many buildds due to RAM exhaustion

[Dejan Latinovic]

Control: tags -1 + patch


Hello,


I was able to avoid this error and successfully build seqan on both mips and 
mipsel,
adding flags -O0 -mxgot.


-O0 was added to solve issue:
  virtual memory exhausted: Cannot allocate memory
during compilation of pair_align.cpp.

-mxgot was added to lift the usual restrictions on the size of the global 
offset table on MIPS,
and to avoid an error:
 relocation truncated to fit: R_MIPS_GOT16
during linking of pair_align.



Patch that contains these changes is attached.
This patch affects only mips and mipsel architectures.

Could you please consider to include it?


Best Regards,
Dejandiff -uNr seqan-1.4.1.orig/debian/rules seqan-1.4.1/debian/rules
--- seqan-1.4.1.orig/debian/rules	2014-01-29 01:57:47.0 +
+++ seqan-1.4.1/debian/rules	2014-07-30 14:09:36.0 +
@@ -8,6 +8,11 @@
 
 DEB_HOST_ARCH   ?= $(shell dpkg-architecture -qDEB_HOST_ARCH)
 
+DEB_BUILD_ARCH  ?= $(shell dpkg-architecture -qDEB_BUILD_ARCH)
+ifneq (,$(filter $(DEB_BUILD_ARCH),mips mipsel))
+export CXXFLAGS:=$(shell dpkg-buildflags --get CXXFLAGS | sed 's/-O[1-9]//') -O0 -mxgot
+endif
+
 %:
 	dh $@
 

Bug#754664: tophat: FTBFS on mips/mipsel: error: invalid abstract return type 'seqan::FileFormat…'

[Dejan Latinovic]

Control: block -1 by 741288

Hello,
this problem is caused using an older version of seqan, 1.3.1-1.
With a new seqan (1.4.1-3), built with proposed patch on #741288,
I was able to build tophat successfully on both mips and mipsel.


Best Regards,
Dejan


--
To UNSUBSCRIBE, email to debian-bugs-rc-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org

Bug#755561: berusky2: FTBFS on mips*: undefined reference to `mmalloc'

[Dejan Latinovic]

Hello,

I have investigated this a little bit more
This seems not to be a compile issue.

The reason for build failure on mips/mipsel is
improper usage of inline functions.

For example in a file:
src/age/graph/mesh.cpp
a functions mmalloc is used.
Declaration of mmalloc function is included from 
src/age/utils/mem_alloc.h,
but a definition of the mentioned function
is in src/komat/mmalloc.h.

mmalloc.h is not visible to mesh.cpp, and during compilation of mesh.cpp
it is not possible to inline mmalloc function.
So, object file  mesh.o is created, but implementation of mmalloc function is 
missing.
This is the reason of an error:
mesh.cpp:59: undefined reference to `mmalloc'
durign linking of berusky2on mips/mipsel with gcc-4.9.

Using gcc-4.8 on mips, in some other object file (in this case 
Berusky3d_light.o)
it seems that compiler decided not to inline mmalloc function.
As a result, during creating of berusky2,
linker is capable to resolve mmalloc symbol,
because Berusky3d_light.o contains implementation of mmalloc function.
In ./komat/Berusky3d_light.cpp definition of mmalloc function
is visible from src/komat/mmalloc.h


The same situation happens on amd64 with gcc-4.9,
but in other object file: Keyframe.o.


The point is that we could not guarantee that inline function
mmalloc would not be inlined in some cases.
For mesh.cpp which is not aware of mmalloc definition,
implementation of mmalloc during linking is needed.

Of course,
we could use -fno-inline flag but, this seems to be wrong solution to me.


The same usage of mmalloc function without a definition visible,
is noticed in files:
age/utils/mem_alloc.cpp
age/graph/gext.cpp
age/utils/utils.cpp


The solution could be to include mmalloc.h in mem_alloc.h,
so the definition of mmalloc function will be visible
in all files where mmalloc function is used.


The patch that contains this solution is attached.
Could you please consider to include it?


Best Regards,
Dejan
diff -uNr berusky2-0.10.orig/src/komat/mmalloc.h berusky2-0.10/src/komat/mmalloc.h
--- berusky2-0.10.orig/src/komat/mmalloc.h	2013-03-03 23:38:39.0 +
+++ berusky2-0.10/src/komat/mmalloc.h	2014-07-24 12:22:36.0 +
@@ -32,6 +32,7 @@
 #include stdlib.h
 #include string.h
 #include assert.h
+#include mmalloc.h
 
 #ifdef __cplusplus
 extern C

Bug#755561: berusky2: FTBFS on mips*: undefined reference to `mmalloc'

[Dejan Latinovic]

Sorry,
i sent a wrong patch.
Here is a good one.

In previous patch I accidentally included mmalloc.h in mmalloc.h :)diff -uNr berusky2-0.10.orig/src/age/utils/mem_alloc.h berusky2-0.10/src/age/utils/mem_alloc.h
--- berusky2-0.10.orig/src/age/utils/mem_alloc.h	2013-03-03 23:38:40.0 +
+++ berusky2-0.10/src/age/utils/mem_alloc.h	2014-07-24 13:35:08.0 +
@@ -36,6 +36,7 @@
 #include unistd.h
 #include limits.h
 #include log.h
+#include mmalloc.h
 
 #define MMALLOC_DEBUG 1
 

Bug#754751: chealpix: FTBFS on mips: E: Caught signal ‘Terminated’: terminating immediately

[Dejan Latinovic]

Hi,

it seams that build of chealpix fails only on Cavium machines.
Even on Cavium, tests execute successfully but
it takes a lot of time.

On Broadcom and Loongson boards,
chealpix builds successfully for me.

The same conclusion could be made from
build logs:
https://buildd.debian.org/status/logs.php?pkg=chealpix

Could we ask for a blacklist chealpix on Cavium boards
corelli, gabrielli and lucatelli?
And ask for a give-back for mips?


Best Regards,
Dejan


--
To UNSUBSCRIBE, email to debian-bugs-rc-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org

Bug#755561: berusky2: FTBFS on mips*: undefined reference to `mmalloc'

[Dejan Latinovic]

Hi,
I test it with gcc/g++ 4.8 on local board,
and berusky2 builds successfully.

Also,
I tried to build an older version of berusky2  (0.10-1),
it fails with g++-4.9 and
builds fine with g++-4.8.

Still,
I did not yet figured out the reason of this behavior.


--
To UNSUBSCRIBE, email to debian-bugs-rc-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org

Bug#754754: jq: FTBFS on s390x: test suite timeout

[Dejan Latinovic]

Hello,
the same error is noticed for other big endian architectures like
mips and powerpc.

This patch from upstream fixes this issue for me on mips.

https://github.com/stedolan/jq/commit/20e629284f94e0c8a14e8425309c913b6ee5b280


The patch is attached.
Could you please consider including this patch.


Best Regards,
Dejandiff -uNr jq-1.4.orig/configure.ac jq-1.4/configure.ac
--- jq-1.4.orig/configure.ac	2014-07-18 15:29:28.0 +
+++ jq-1.4/configure.ac	2014-07-18 14:37:53.0 +
@@ -96,6 +96,12 @@
 fi
 AC_MSG_RESULT($have___thread)
 
+AC_C_BIGENDIAN(
+   AC_DEFINE([IEEE_MC68k], 1, [machine is bigendian]),
+   AC_DEFINE([IEEE_8087], 1, [machine is littleendian]),
+   AC_MSG_ERROR(unknown endianess),
+   AC_MSG_ERROR(universial endianess not supported)
+)
 
 AC_SUBST([BUNDLER], [$bundle_cmd])
 
diff -uNr jq-1.4.orig/jv_dtoa.c jq-1.4/jv_dtoa.c
--- jq-1.4.orig/jv_dtoa.c	2014-06-19 23:31:18.0 +
+++ jq-1.4/jv_dtoa.c	2014-07-18 14:38:17.0 +
@@ -187,7 +187,6 @@
  *	used for input more than STRTOD_DIGLIM digits long (default 40).
  */
 
-#define IEEE_8087
 #define NO_ERRNO
 #define NO_HEX_FP
 #define No_Hex_NaN

Bug#751343: python-tornado: FTBFS on mips due to test failure

[Dejan Latinovic]

Hello Scott,

It seams that this failure
is related to board performance,
since this test failed three times but always on corelli (Cavium).

I was not able to reproduce this test failure on my local boards,
aside from decreasing time limit below 1.0.

To resolve this issue we could increase time limit to 2.0 or 1.6,
since time of test_multi_performance execution on corelli
is between 1.4046919345855713 and 1.5510458946228027,
at least according to existing build logs.

The other solution could be to blacklist python-tornado
on corelli.


Best Regards,
Dejan

--
To UNSUBSCRIBE, email to debian-bugs-rc-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org

Bug#743685: urwid: FTBFS: test failures

[Dejan Latinovic]

Debian patch that removes test_vterm.py
is attached.

Vincent, could you please include this patch 
in new Debian package version?


Regards,
Dejan



diff -uNr urwid-1.2.1.orig/urwid/tests/test_vterm.py urwid-1.2.1/urwid/tests/test_vterm.py
--- urwid-1.2.1.orig/urwid/tests/test_vterm.py	2014-04-04 15:55:01.0 +
+++ urwid-1.2.1/urwid/tests/test_vterm.py	1970-01-01 00:00:00.0 +
@@ -1,334 +0,0 @@
-# Urwid terminal emulation widget unit tests
-#Copyright (C) 2010  aszlig
-#Copyright (C) 2011  Ian Ward
-#
-#This library is free software; you can redistribute it and/or
-#modify it under the terms of the GNU Lesser General Public
-#License as published by the Free Software Foundation; either
-#version 2.1 of the License, or (at your option) any later version.
-#
-#This library is distributed in the hope that it will be useful,
-#but WITHOUT ANY WARRANTY; without even the implied warranty of
-#MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-#Lesser General Public License for more details.
-#
-#You should have received a copy of the GNU Lesser General Public
-#License along with this library; if not, write to the Free Software
-#Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
-#
-# Urwid web site: http://excess.org/urwid/
-
-import os
-import sys
-import unittest
-
-from itertools import dropwhile
-
-from urwid import vterm
-from urwid import signals
-from urwid.compat import B
-
-
-class DummyCommand(object):
-QUITSTRING = B('|||quit|||')
-
-def __init__(self):
-self.reader, self.writer = os.pipe()
-
-def __call__(self):
-# reset
-stdout = getattr(sys.stdout, 'buffer', sys.stdout)
-stdout.write(B('\x1bc'))
-
-while True:
-data = os.read(self.reader, 1024)
-if self.QUITSTRING == data:
-break
-stdout.write(data)
-stdout.flush()
-
-def write(self, data):
-os.write(self.writer, data)
-
-def quit(self):
-self.write(self.QUITSTRING)
-
-
-class TermTest(unittest.TestCase):
-def setUp(self):
-self.command = DummyCommand()
-
-self.term = vterm.Terminal(self.command)
-self.resize(80, 24)
-
-def tearDown(self):
-self.command.quit()
-
-def connect_signal(self, signal):
-self._sig_response = None
-
-def _set_signal_response(widget, *args, **kwargs):
-self._sig_response = (args, kwargs)
-self._set_signal_response = _set_signal_response
-
-signals.connect_signal(self.term, signal, self._set_signal_response)
-
-def expect_signal(self, *args, **kwargs):
-self.assertEqual(self._sig_response, (args, kwargs))
-
-def disconnect_signal(self, signal):
-signals.disconnect_signal(self.term, signal, self._set_signal_response)
-
-def caught_beep(self, obj):
-self.beeped = True
-
-def resize(self, width, height, soft=False):
-self.termsize = (width, height)
-if not soft:
-self.term.render(self.termsize, focus=False)
-
-def write(self, data):
-data = B(data)
-self.command.write(data.replace(B('\e'), B('\x1b')))
-
-def flush(self):
-self.write(chr(0x7f))
-
-def read(self, raw=False):
-self.term.wait_and_feed()
-rendered = self.term.render(self.termsize, focus=False)
-if raw:
-is_empty = lambda c: c == (None, None, B(' '))
-content = list(rendered.content())
-lines = [list(dropwhile(is_empty, reversed(line)))
- for line in content]
-return [list(reversed(line)) for line in lines if len(line)]
-else:
-content = rendered.text
-lines = [line.rstrip() for line in content]
-return B('\n').join(lines).rstrip()
-
-def expect(self, what, desc=None, raw=False):
-if not isinstance(what, list):
-what = B(what)
-got = self.read(raw=raw)
-if desc is None:
-desc = ''
-else:
-desc += '\n'
-desc += 'Expected:\n%r\nGot:\n%r' % (what, got)
-self.assertEqual(got, what, desc)
-
-def test_simplestring(self):
-self.write('hello world')
-self.expect('hello world')
-
-def test_linefeed(self):
-self.write('hello\x0aworld')
-self.expect('hello\nworld')
-
-def test_linefeed2(self):
-self.write('aa\b\b\eDbb')
-self.expect('aa\nbb')
-
-def test_carriage_return(self):
-self.write('hello\x0dworld')
-self.expect('world')
-
-def test_insertlines(self):
-self.write('\e[0;0flast\e[0;0f\e[10L\e[0;0ffirst\nsecond\n\e[11D')
-self.expect('first\nsecond\n\n\n\n\n\n\n\n\nlast')
-
-def test_deletelines(self):
-self.write('1\n2\n3\n4\e[2;1f\e[2M')
-self.expect('1\n4')
-
-def test_movement(self):
-  

Bug#741743: kicad: FTBFS: find: `debian/kicad-common/usr/share/kicad': No such file or directory

[Dejan Latinovic]

Hi,


It seams to me that the problem is in debian/rules file.
If package is being built with both arch specific and indep parts (fakeroot 
debian/rules binary),
build finishes successfully .
If we build only arch specific parts of the package (binary-arch),
build fails on:
 find debian/kicad-common/usr/share/kicad -name CMake* -exec rm {} \;
 find: `debian/kicad-common/usr/share/kicad': No such file or directory

The reason for this is the fact that kicad-common is arch independent,
and should not be packed in this case.

In order to resolve this issue I split override_dh_install into
two parts: 
override_dh_install-arch and override_dh_install-indep.

Changes needed for debian/rules are attached.

With this fix I was able to successfully build kicad for
mips, mipsel and amd64.


Regards,
Dejan

--- kicad-0.20140224+bzr4027.orig/debian/rules	2014-03-07 10:50:28.0 +
+++ kicad-0.20140224+bzr4027/debian/rules	2014-05-08 14:38:34.0 +
@@ -21,7 +21,7 @@
 override_dh_compress:
 	dh_compress --exclude=.pdf
 
-override_dh_install:
+override_dh_install-arch:
 	convert bitmaps_png/icons/icon_kicad.xpm -resize 32x32 debian/tmp/icon_kicad.xpm
 	dh_install
 	# fix wrong permissions and link jnlp file to usr/share
@@ -29,6 +29,9 @@
 	  chmod a-x $(COMMONDIR)/usr/share/kicad/internat/ja/*
 	ln -s ../share/kicad/freeroute.jnlp $(PKGDIR)/usr/bin
 	chmod a-x $(PKGDIR)/usr/share/kicad/freeroute.jnlp
+
+override_dh_install-indep:
+	dh_install
 	# remove out-of-topic files
 	find $(COMMONDIR)/usr/share/kicad -name CMake* -exec rm {} \;
 

Bug#599972: argus: FTBFS on mips*: cp: cannot stat `./bin/argus': No such file or directory

[Dejan Latinovic]

Hello,


Instead of statically linking,
we could replace the private libpcap function pcap_offline_read
with the proper pcap_dispatch function.


A similar issue is reported here:
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=545595



The patch that include this change is attached.



Regards,
Dejan
--- argus-3.0.0.orig/argus/ArgusSource.c	2008-03-17 15:00:51.0 +
+++ argus-3.0.0/argus/ArgusSource.c	2014-04-11 12:12:36.0 +
@@ -2209,7 +2209,7 @@
   } else {
  for (i = 0; i  src-ArgusInterfaces; i++) {
 src-ArgusThisIndex = i;
-pcap_offline_read (src-ArgusInterface[i].ArgusPd, src-eNflag,
+pcap_dispatch (src-ArgusInterface[i].ArgusPd, src-eNflag,
src-ArgusInterface[i].ArgusCallBack, (u_char *) src);
  }
   }

Bug#736565: [Debian-med-packaging] Bug#736565: FTBFS on non-PC architectures: FAIL: scram_mt.test

[Dejan Latinovic]

Hello,

we have been investigating this issue.

We will have a patch that fixes the issue in a day or two.


Regards,
Dejan



From: Anibal Monsalve Salazar [ani...@master.debian.org] on behalf of Aníbal 
Monsalve Salazar [ani...@debian.org]
Sent: Monday, March 31, 2014 10:26 AM
To: Charles Plessy
Cc: 736...@bugs.debian.org; Dejan Latinovic; debian-m...@lists.debian.org
Subject: Re: Bug#736565: [Debian-med-packaging] Bug#736565: FTBFS on non-PC 
architectures: FAIL: scram_mt.test

On Mon, 2014-03-31 18:03:12 +1100, Aníbal Monsalve Salazar wrote:
 On Mon, 2014-03-31 14:06:49 +0900, Charles Plessy wrote:

 staden-io-lib is going to be removed from Testing because of a RC bug
 that is solved in Sid but can not migrate as long as the package still
 fails to build o non-PC architectures.

 My original plan was to remove staden-io-lib from [armel armhf mips
 mipsel powerpc s390x ia64], but this was cancelled as you voluntered
 to fix the build failure (http://bugs.debian.org/736565#10).

 Do you know when you will be able to fix the package ?  If it is not
 likely to happen in the short term, I would like to follow my original
 plan and remove it from the architectures where it does not build.

 We have staden-io-lib in a our to-do list. We haven't forgotten it.

 diff -up ./tmp/buildd/staden-io-lib-1.13.3/tests/test.out/xx#minimal.full.sam 
 ./tmp/buildd/staden-io-lib-1.13.3/tests/test.out/xx#minimal.full.bam.sam
 --- ./tmp/buildd/staden-io-lib-1.13.3/tests/test.out/xx#minimal.full.sam  
   2014-03-09 09:28:59.776817773 +
 +++ ./tmp/buildd/staden-io-lib-1.13.3/tests/test.out/xx#minimal.full.bam.sam  
   2014-03-09 09:29:00.120803439 +
 @@ -10,4 +10,3 @@ A016  yy  4   1   *   *   0 
   0   *   *
  A1 16  yy  4   1   *   *   0   0   * 
   *
  A2 16  yy  4   1   *   *   0   0   * 
   *
  A3 16  yy  4   1   *   *   0   0   * 
   *
 -A4 16  yy  4   1   *   *   0   0   * 
   *

 The missing line in xx#minimal.full.bam.sam is still causing the failure
 with the new version 1.13.3.

Version 1.13.5.

I was looking at the wrong package version.

http://people.debian.org/~anibal/staden-io-lib/staden-io-lib_1.13.5-1_mipsel.log.bz2

The build log extract is below. The web address of the complete build
log is above.

make[2]: Leaving directory `/tmp/buildd/staden-io-lib-1.13.5/tests'
make[1]: *** [check-recursive] Error 1
make[1]: Leaving directory `/tmp/buildd/staden-io-lib-1.13.5'
dh_auto_test: make -j1 check returned exit code 2
make: *** [build] Error 2
dpkg-buildpackage: error: debian/rules build gave error exit status 2
E: Failed autobuilding of package
I: user script /var/cache/pbuilder/build//30984/tmp/hooks/C10cmds starting
+ find -name xx#minimal.full.sam
+ find -name xx#minimal.full.bam.sam
+ diff -up ./tmp/buildd/staden-io-lib-1.13.5/tests/test.out/xx#minimal.full.sam 
./tmp/buildd/staden-io-lib-1.13.5/tests/test.out/xx#minimal.full.bam.sam
--- ./tmp/buildd/staden-io-lib-1.13.5/tests/test.out/xx#minimal.full.sam
2014-03-31 07:56:46.008964719 +
+++ ./tmp/buildd/staden-io-lib-1.13.5/tests/test.out/xx#minimal.full.bam.sam
2014-03-31 07:56:46.348950575 +
@@ -10,4 +10,3 @@ A016  yy  4   1   *   *   0   
0   *   *
 A1 16  yy  4   1   *   *   0   0   *   
*
 A2 16  yy  4   1   *   *   0   0   *   
*
 A3 16  yy  4   1   *   *   0   0   *   
*
-A4 16  yy  4   1   *   *   0   0   *   
*
I: unmounting dev/pts filesystem
I: unmounting run/shm filesystem
I: unmounting proc filesystem
I: cleaning the build env
I: removing directory /var/cache/pbuilder/build//30984 and its subdirectories


 Do you have the upstream email address?

 I tried to email usptream and the email address in debian/copyright
 wasn't that of upstream.

--
To UNSUBSCRIBE, email to debian-bugs-rc-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org

Bug#711787: falconpl: FTBFS on mipsen

[Dejan Latinovic]

Hi,

Falcon has it own basic memory manager functions.
In these functions, during memory allocation,
the size of allocated memory is expanded by sizeof(size_t)
[4 bytes, on 32-bit MIPS],
which is used for saving an amount of allocated memory.

The first 4 bytes are used to store the amount.
Because of that,
the pointer is offset by these 4 bytes and
address of the actual data is not aligned by 8.

Bus Error happens for data with an amount of 8 bytes
or larger, due to usage of 32-bit MIPS
instructions ldc1 and sdc1,
for double type access on  address not aligned by 8.


To avoid this behavior I add one more
block of 4 bytes (sizeof(size_t)) during allocation,
to realign the data address.


With these changes I was able to built the falconpl successfully
on my local MIPS board.

Besides that,
I run tests from tests/core/testsuite/.
There were no fails with a Bus Error.


The patch that contains these changes is attached.

Could you consider including this patch?


Thanks,
Dejan
--- a/engine/memory.cpp
+++ b/engine/memory.cpp
@@ -70,7 +70,11 @@
 
 void * DflAccountMemAlloc( size_t amount )
 {
+#if defined (__mips__ )  !defined(__mips64)
+   size_t *ret =  (size_t*) malloc( amount + 2 * sizeof(size_t) );
+#else
size_t *ret =  (size_t*) malloc( amount + sizeof(size_t) );
+#endif
if ( ret == 0 ) {
   printf( Falcon: fatal allocation error when allocating %d bytes\n, (int) amount );
   exit(1);
@@ -78,7 +82,11 @@
 
gcMemAccount( amount );
*ret = amount;
+#if defined (__mips__ )  !defined(__mips64)
+   return ret+2;
+#else
return ret+1;
+#endif
 }
 
 
@@ -87,8 +95,13 @@
if ( mem != 0 )
{
   size_t *smem = (size_t*) mem;
+#if defined (__mips__ )  !defined(__mips64)
+  gcMemUnaccount( smem[-2] );
+  free( smem-2 );
+#else
   gcMemUnaccount( smem[-1] );
   free( smem-1 );
+#endif
}
 }
 
@@ -105,10 +118,18 @@
 
 
size_t *smem = (size_t*) mem;
+#if defined (__mips__ )  !defined(__mips64)
+   smem-=2;
+#else
smem--;
+#endif
size_t oldalloc = *smem;
 
+#if defined (__mips__ )  !defined(__mips64)
+   size_t *nsmem = (size_t*) realloc( smem, amount + 2 * sizeof( size_t ) );
+#else
size_t *nsmem = (size_t*) realloc( smem, amount + sizeof( size_t ) );
+#endif
 
if ( nsmem == 0 ) {
   printf( Falcon: fatal reallocation error when allocating %d bytes\n, (int) amount );
@@ -121,7 +142,11 @@
else
   gcMemUnaccount( oldalloc - amount );
 
+#if defined (__mips__ )  !defined(__mips64)
+   return nsmem+2;
+#else
return nsmem+1;
+#endif
 }
 
 

Bug#736481: nettoe: FTBFS due to test suite failures

[Dejan Latinovic]

Hello,

It seems that libncurses5-dev is added as build dependency
at latest version (1.5-1) of nettoe.
Due to that, libncurses5-dev and libtinfo-dev
were installed.
On configuration, term.h and libtermcap.so
were found and it is guessed that TERM environment variable is defined.
That was not the case in previous version (1.4.2-1),
so nettoe has been built successfully.

I have not figured out why TERM is defined on some 
build machines, and on others is not.

However, it is possible to build nettoe
without terminfo if TERM environment variable
is not defined.

A patch that contains these changes is attached.

Could you consider applying this patch?


Regards,
Dejan
--- nettoe-1.5.orig/debian/rules	2013-11-29 16:26:18.0 +
+++ nettoe-1.5/debian/rules	2014-03-19 17:34:36.0 +
@@ -4,11 +4,16 @@
 
 include /usr/share/dpkg/buildflags.mk
 
+ifndef TERM
+CONFIG_EXTRA=--without-terminfo
+endif
+
 %:
 	dh $@ --with autoreconf
 
 override_dh_auto_configure:
-	./configure --prefix=/usr \
+	./configure $(CONFIG_EXTRA) \
+		--prefix=/usr \
 		--bindir=/usr/games \
 		--mandir=/usr/share/man \
 		--enable-desktop

Bug#720833: eegdev: FTBFS: conflicting types for 'cff_parse'

[Dejan Latinovic]

Hello,

The reason for that is a migration to the newer version of bison,
from 2.5 to 3.0.2.

Attached patch contains changes to bison input files (src/core/conffile.y and 
src/core/confline.y),
in order to adapt to newer bison version.

Could you please consider applying this patch?

Thanks,
Dejan
--- a/src/core/conffile.y
+++ b/src/core/conffile.y
@@ -17,9 +17,11 @@
 along with this program.  If not, see http://www.gnu.org/licenses/.
 */
 %define api.pure
+%no-lines
 %name-prefix cff_
 %defines conffile.tab.h
 %parse-param { struct cfdata *pp }
+%lex-param { yyscan_t cff_scaninfo }
 %{
 #if HAVE_CONFIG_H
 # include config.h
@@ -28,6 +30,11 @@
 #include stdio.h
 #include stdlib.h
 #include configuration.h
+#ifndef YY_TYPEDEF_YY_SCANNER_T
+#define YY_TYPEDEF_YY_SCANNER_T
+typedef void* yyscan_t;
+#endif
+#include conffile.h
 %}
 
 %union value {
@@ -36,9 +43,7 @@
 
 %{
 #include conffile.lex.h
-#include conffile.h
-#define YYLEX_PARAM pp-scaninfo
-
+#define cff_scaninfo pp-scaninfo
 static int yyerror(struct cfdata *pp, const char* s);
 %}
 
--- a/src/core/confline.y
+++ b/src/core/confline.y
@@ -17,9 +17,11 @@
 along with this program.  If not, see http://www.gnu.org/licenses/.
 */
 %define api.pure
+%no-lines
 %name-prefix cfl_
 %defines confline.tab.h
 %parse-param { struct cfldata *pp }
+%lex-param { yyscan_t cfl_scaninfo }
 %{
 #if HAVE_CONFIG_H
 # include config.h
@@ -28,6 +30,11 @@
 #include stdio.h
 #include stdlib.h
 #include configuration.h
+#ifndef YY_TYPEDEF_YY_SCANNER_T
+#define YY_TYPEDEF_YY_SCANNER_T
+typedef void* yyscan_t;
+#endif
+#include confline.h
 %}
 
 %union value {
@@ -36,9 +43,7 @@
 
 %{
 #include confline.lex.h
-#include confline.h
-#define YYLEX_PARAM pp-scaninfo
-
+#define cfl_scaninfo pp-scaninfo
 static int yyerror(struct cfldata *pp, const char* s);
 %}
 

Bug#740985: FTBFS on amd64 mips mipsel: checking for KQUEUE... configure: error: Package requirements (libkqueue) were not met

[Dejan Latinovic]

Hello,

It is not a problem to locate libkqueue.pc,
but content of .pc file.

libkqueue.pc in 2.0.2-1:

Name: libkqueue
Description: Emulates FreeBSD kqueue(2) on other platforms
Version: 2.0.2
Requires:
Libs: -lkqueue
Libs.private: -lpthread -lrt
Cflags: -I${includedir}/kqueue


libkqueue.pc in 1.0.4-2:

prefix=/usr
exec_prefix=${prefix}
libdir=/usr/lib
includedir=/usr/include
Name: libkqueue
Description: Emulates FreeBSD kqueue(2) on other platforms
Version: 1.0.4
Requires:
Libs:  -L/usr/lib -lpthread -lrt -lkqueue
Libs.private:  -L/usr/lib -lpthread -lrt
Cflags: -I${includedir}/kqueue



I believe that problem is caused by changes in libkqueue.pc.in:

--- libkqueue-1.0.4/libkqueue.pc.in 2011-07-14 01:48:14.0 +
+++ libkqueue-2.0.2/libkqueue.pc.in 2014-02-23 21:52:26.0 +
@@ -1,13 +1,8 @@
-prefix=@@PREFIX@@
-exec_prefix=${prefix}
-libdir=@@LIBDIR@@
-includedir=@@INCLUDEDIR@@
 
-Name: @@PROGRAM@@
+Name: libkqueue
 Description: Emulates FreeBSD kqueue(2) on other platforms
-Version: @@VERSION@@
+Version: @VERSION@
 Requires: 
-Libs: @@LIBDEPENDS@@ -lkqueue 
-Libs.private: @@LIBDEPENDS@@
-Cflags: -I${includedir}/kqueue 
-
+Libs: -lkqueue
+Libs.private: -lpthread -lrt
+Cflags: -I${includedir}/kqueue


It seems to me that better solution is to fix libkqueue
in order to prevent the error propagates to other packages.



Regards,
Dejan

--
To UNSUBSCRIBE, email to debian-bugs-rc-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org

Bug#740486: opendkim: FTBFS on mipsen (test failures)

[Dejan Latinovic]

Hello,

As I see from debian/changelog,
tests for all architectures should be disabled.

Snip from changelog:
 opendkim (2.9.0-2) unstable; urgency=low
 
   * Disable test suite on all archs due to undiagnosed race condition in the
 test suite that causes multiple FTBFS, but doesn't appear reflect any
 actual program problems
 
  -- Scott Kitterman sc...@kitterman.com  Sat, 08 Feb 2014 23:22:17 -0500

It seems that Scott missed to disable tests for mips/mipsel.

A patch that disables tests for mips/misel
is attached.


Regards,
Dejan--- opendkim-2.9.0.orig/debian/rules	2014-03-05 17:13:43.0 +
+++ opendkim-2.9.0/debian/rules	2014-03-05 17:13:23.0 +
@@ -33,10 +33,10 @@
 		--with-test-socket=inet:8891@localhost \
 		--with-lua
 
-ifneq (,$(filter $(DEB_HOST_ARCH), mips mipsel))
-override_dh_auto_test:
-	MILTERTEST_RETRY_SPEED_FACTOR=20 dh_auto_test
-endif
+#ifneq (,$(filter $(DEB_HOST_ARCH), mips mipsel))
+#override_dh_auto_test:
+#	MILTERTEST_RETRY_SPEED_FACTOR=20 dh_auto_test
+#endif
 
 # test results on kfreebsd-* are inconsistent, but not due to package bugs.
 #ifneq (,$(filter $(DEB_HOST_ARCH), kfreebsd-i386 kfreebsd-amd64))

Bug#738137: libpwiz: FTBFS on mips*: virtual memory exhausted

[Dejan Latinovic]

Hello,

I have attached a patch that solves
this issue for me on mips/mipsel.

Regards,
Dejan

--- libpwiz-3.0.4624.orig/debian/rules	2014-02-11 14:46:15.0 +
+++ libpwiz-3.0.4624/debian/rules	2014-02-27 09:58:59.0 +
@@ -12,6 +12,7 @@
 
 DEB_HOST_GNU_TYPE	?= $(shell dpkg-architecture -qDEB_HOST_GNU_TYPE)
 DEB_BUILD_GNU_TYPE	?= $(shell dpkg-architecture -qDEB_BUILD_GNU_TYPE)
+DEB_BUILD_ARCH  ?= $(shell dpkg-architecture -qDEB_BUILD_ARCH)
 
 DEBIAN_DIR = $(CURDIR)/debian
 BUILD_DIR = $(DEBIAN_DIR)/build
@@ -70,6 +71,10 @@
 CFLAGS+=$(CPPFLAGS)
 CXXFLAGS+=$(CPPFLAGS)
 
+ifneq (,$(filter $(DEB_BUILD_ARCH),mips mipsel))
+   CXXFLAGS += --param ggc-min-expand=5
+endif
+
 .PHONY: clean
 clean: 
 # unpatch

Bug#739807: FTBFS: error: 'uintptr_t' was not declared in this scope

[Dejan Latinovic]

Hello,

I was getting the same error with libpwiz packge.

 checking for boost/system/error_code.hpp... yes
 checking for the Boost system library... yes
 checking boost/thread.hpp usability... yes
 checking boost/thread.hpp presence... yes
 checking for boost/thread.hpp... yes
 configure: error: cannot compile a test that uses Boost thread
 make: *** [build-arch-stamp] Error 1

Tail from config.log:

 configure:25276: checking for the Boost thread library
 configure:25330: g++ -c -g -O2 -fstack-protector --param=ssp-buffer-size=4 
 -Wformat -Werror=format-security -D_FORTIFY_SOURCE=2 -D_FORTIFY_SOURCE=2 
 -pthread  conftest.cpp 5
 In file included from /usr/include/boost/atomic.hpp:12:0,
  from /usr/include/boost/thread/pthread/once_atomic.hpp:20,
  from /usr/include/boost/thread/once.hpp:20,
  from /usr/include/boost/thread.hpp:17,
  from conftest.cpp:63:
 /usr/include/boost/atomic/atomic.hpp:202:16: error: 'uintptr_t' was not 
 declared in this scope
  typedef atomicuintptr_t atomic_uintptr_t;
 ^
 /usr/include/boost/atomic/atomic.hpp:202:25: error: template argument 1 is 
 invalid
  typedef atomicuintptr_t atomic_uintptr_t;
  ^
 /usr/include/boost/atomic/atomic.hpp:202:43: error: invalid type in 
 declaration before ';' token
  typedef atomicuintptr_t atomic_uintptr_t;


After I built boost1.54 with patch from
https://svn.boost.org/trac/boost/changeset/84950,
I was able to build libpwiz successfully.


Regards,
Dejan

--
To UNSUBSCRIBE, email to debian-bugs-rc-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org

Bug#739184: ray: FTBFS on mips/mipsel/s390x: error: 'memcpy' was not declared in this scope

[Dejan Latinovic]

Hello,

Apart form mips/mipsel architectures, the declaration of memcpy is reached 
through openmpi.
Since the mpi-defaults for mips/mipsel was build against mpich (not openmpi),
this issue appeared.
Therefore, package ray does build succsessfully on other architectures without 
changes.

New mpi-defaults version 1.0.2+nmu1 is build against openmpi for mips/mipsel.

--
To UNSUBSCRIBE, email to debian-bugs-rc-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org

Bug#731967: FW: Bug#731967: libqb FTBFS: build failed on post-compile-test on mips/mipsel

[Dejan Latinovic]

Hi,


The bug is fixed upstream.
Can you consider getting the latest source from
upstream?

https://github.com/ClusterLabs/libqb


Regards,
Dejan LatinovićAuthor: Dejan Latinovic dejan.latino...@rt-rk.com
Description: Check if call of function will go through lazy binding stub for mips.
Index: patch.gammaray-1.2.2/launcher/injector/preloadcheck.cpp
===
--- patch.gammaray-1.2.2.orig/launcher/injector/preloadcheck.cpp	2012-12-21 21:38:15.0 +
+++ patch.gammaray-1.2.2/launcher/injector/preloadcheck.cpp	2013-11-27 16:31:07.0 +
@@ -99,10 +99,99 @@
 }
   }
 
+//Mips, besides the plt, has another method of
+//calling functions from .so files, and this method doesn't need JUMP_SLOT
+//relocations (in fact, it doesn't need any relocations). This method uses .got
+//entries and lazy binding stubs.
+#ifdef __mips__
+  if (testMips(symbol, fileName)){
+qDebug()  Call of function   symbol   will go through lazy binding stub;
+setErrorString(QString());
+return true;
+  }
+#endif
+
   setErrorString(QObject::tr(Symbol is not marked as relocatable: %1).arg(symbol));
   return false;
 }
 
+
+//The way to determine whether the call to function will go
+//through .got and lazy binding stub is:
+//- find the value of dynamic symbol index of the function with the command
+//  readelf --dyn-syms
+//- find the value of dynamic tag MIPS_GOTSYM with the command readelf -d
+//- if (dyn_sym_index = MIPS_GOTSYM) then the function has entry in the global
+//  part of the .got, and the call will go through lazy binding stub and be
+//  resolved by dynamic linker.
+
+#ifdef __mips__
+bool PreloadCheck::testMips(const QString symbol, const QString fileName)
+{
+  QProcess proc;
+  proc.setProcessChannelMode(QProcess::MergedChannels);
+  proc.start(readelf, QStringList()  --dyn-syms  -d  fileName, QIODevice::ReadOnly);
+
+  if (!proc.waitForFinished()) {
+setErrorString(QObject::tr(Failed to run 'readelf' (binutils) binary: %1).
+   arg(QString(proc.errorString(;
+return false;
+  }
+
+  if (proc.exitCode() != 0) {
+setErrorString(QObject::tr(Cannot read shared object: %1).arg(QString(proc.readAll(;
+return false;
+  }
+
+  //Example line of dynamic symbol table on mips:
+  //3851: 001e66f4 8 FUNCGLOBAL DEFAULT   11 qt_startup_hook
+  QRegExp rxSym(^(\\d+):\\s+(?:[^ ]+\\s+){6}([^ ]+)(?:.*)$);
+
+  //Example line of dynamic tag on mips:
+  //0x7013 (MIPS_GOTSYM)0xec3
+  QRegExp rxGot(^0x[0-9a-fA-F]+\\s+\\((.+)\\)\\s+(0x[0-9a-fA-F]+)(?:.*)$);
+
+  int dyn_sym_index = 0;
+  int mips_gotsym = 0;
+  bool foundDynSymbol = false;
+  bool foundGotTag = false;
+
+  while (proc.canReadLine()) {
+const QString line = proc.readLine().trimmed();
+QString currentMatch;
+const QString tag = MIPS_GOTSYM;
+
+if (rxGot.exactMatch(line)) {
+  currentMatch = rxGot.cap(1);
+
+  if (currentMatch == tag) {
+bool conversionOk = false;
+int value = rxGot.cap(2).toInt(conversionOk, 16);
+if (conversionOk){
+  mips_gotsym = value;
+  foundGotTag = true;
+}
+  }
+}
+else if (rxSym.exactMatch(line)) {
+  currentMatch = rxSym.cap(2);
+  if (currentMatch == symbol) {
+dyn_sym_index = rxSym.cap(1).toInt();
+foundDynSymbol = true;
+  }
+}
+if (foundGotTag  foundDynSymbol)
+  break;
+  }
+  if (foundGotTag  foundDynSymbol  (dyn_sym_index = mips_gotsym)) {
+return true;
+  }
+
+  return false;
+}
+#endif
+
+
 void PreloadCheck::setErrorString(const QString err)
 {
   m_errorString = err;
Index: patch.gammaray-1.2.2/launcher/injector/preloadcheck.h
===
--- patch.gammaray-1.2.2.orig/launcher/injector/preloadcheck.h	2012-12-21 21:38:15.0 +
+++ patch.gammaray-1.2.2/launcher/injector/preloadcheck.h	2013-11-27 16:31:07.0 +
@@ -44,6 +44,10 @@
   private:
 static QString findSharedObjectFile(const QString symbol);
 
+#ifdef __mips__
+bool testMips(const QString symbol, const QString fileName);
+#endif
+
 QString m_errorString;
 };
 

Bug#599972: argus: FTBFS on mips*: cp: cannot stat `./bin/argus': No such file or directory

[Dejan Latinovic]

Control: tags -1 + patch


Hi,

this bug also occurs on other architectures:
  https://buildd.debian.org/status/logs.php?pkg=argussuite=experimental


Snip of build log:
  ArgusSource.o: In function `ArgusGetPackets':
  
/build/buildd-argus_3.0.0-3-mips-ANkqA6/argus-3.0.0/argus/ArgusSource.c:2212: 
undefined reference to `pcap_offline_read'
  collect2: ld returned 1 exit status
  make[2]: *** [../bin/argus] Error 1

The solution is to set libpcap library path.

The patch that solves this problem is attached.
With this patch I successfully built argus for following architectures:  misp, 
mipsel, i386 and amd64.

Regards,
Dejan Latinović--- argus-3.0.0.orig/debian/rules	2014-01-09 17:25:55.0 +
+++ argus-3.0.0/debian/rules	2014-01-09 16:25:25.920618844 +
@@ -14,6 +14,7 @@
 
 export DEB_HOST_GNU_TYPE  ?= $(shell dpkg-architecture -qDEB_HOST_GNU_TYPE)
 export DEB_BUILD_GNU_TYPE ?= $(shell dpkg-architecture -qDEB_BUILD_GNU_TYPE)
+DEB_HOST_MULTIARCH ?= $(shell dpkg-architecture -qDEB_HOST_MULTIARCH)
 
 ifeq ($(DEB_BUILD_GNU_TYPE), $(DEB_HOST_GNU_TYPE))
   confflags += --build $(DEB_HOST_GNU_TYPE)
@@ -42,7 +43,8 @@
 	--bindir=/usr/bin \
 	--sysconfdir=/etc \
 	--mandir=/usr/share/man \
-	--includedir=/usr/include CFLAGS=$(CFLAGS)
+	--includedir=/usr/include CFLAGS=$(CFLAGS) \
+	--with-libpcap=/usr/lib/$(DEB_HOST_MULTIARCH)
 	
 binary-indep: build
 	dh_testroot

Bug#733917: linbox FTBFS: build failed on post-compile-test on mips/mipsel

[Dejan Latinovic]

Package: linbox
Version: 1.3.2-1
Severity: serious
Tags: sid patch
Justification: FTBFS


In an attempt to build libqb on mips/mipsel, 
build failed on testing:


  PASS: test-charpoly
  make[5]: *** Deleting file `test-det.log'
  make[5]: *** [test-det.log] Terminated
  make: *** [build-arch] Terminated
  E: Caught signal ‘Terminated’: terminating immediately
  make[4]: *** [check-TESTS] Terminated
  make[1]: *** [check-recursive] Terminated
  make[2]: *** [check-recursive] Terminated
  make[3]: *** [check-am] Terminated
  Terminated
  Build killed with signal TERM after 300 minutes of inactivity

The full build logs are available from:
https://buildd.debian.org/status/fetch.php?pkg=linboxarch=mipsver=1.3.2-1stamp=1379576905

Test test-det on i386 or amd64 successfully pass after a few iterations,
on mips this test has an enormous number of iterations
and build of the linbox package on buildd breaks after 300 minutes of 
inactivity.

The reason for this behavior is that in:
linbox/algorithms/rational-reconstruction-base.h
the way to calculate RecCounter is:
  RecCounter = (size_t)((double)log((double)rbound_)/log(2.));//2^RecCounter  
rbound_ =2^(RecCounter+1)
For rbound_ is equal to zero the result of log function is -inf (double[8 
bytes]).
When we cast -inf to size_t (unsigned int[4 bytes]) on i386 and amd64,
result is zero.
In the same situation on mips architecture, cast is done through
TRUNC.W.D instruction.
For this instraction, if the source value is Infinity, NaN, or rounds to an 
integer outside the range
-2^31 to 2^31-1, the result cannot be represented correctly and an IEEE Invalid 
Operation
condition exists. The result depends on the FP exception model currently active 
and the
default result is 2^31–1 (2147483647).


A patch fixing this issue is attached.
Author: Dejan Latinovic dejan.latino...@rt-rk.com
Description: Set RecCounter=0 if rbound_==0 to avoid cast of -inf.
Index: linbox-1.3.2/linbox/algorithms/rational-reconstruction-base.h
===
--- linbox-1.3.2.orig/linbox/algorithms/rational-reconstruction-base.h	2012-06-09 02:13:48.0 +
+++ linbox-1.3.2/linbox/algorithms/rational-reconstruction-base.h	2013-12-31 13:40:03.0 +
@@ -73,7 +73,8 @@
 RecCounter = (int)sqrt((double)rbound_);//RecCounter^2  rbound_ =(RecCounter+1)^2
 			}
 			else if (_meth == GEOMETRIC) {
-RecCounter = (size_t) log((double)rbound_) ;//2^RecCounter  rbound_ =2^(RecCounter+1)
+if (rbound_!=0)
+	RecCounter = (size_t) log((double)rbound_) ;//2^RecCounter  rbound_ =2^(RecCounter+1)
 			}
 		}
 
@@ -85,7 +86,8 @@
 RecCounter = (size_t)sqrt((double)rbound_);//RecCounter^2  rbound_ =(RecCounter+1)^2
 			}
 			else if (_meth == GEOMETRIC) {
-RecCounter = (size_t)((double)log((double)rbound_)/log(2.));//2^RecCounter  rbound_ =2^(RecCounter+1)
+if (rbound_!=0)
+	RecCounter = (size_t)((double)log((double)rbound_)/log(2.));//2^RecCounter  rbound_ =2^(RecCounter+1)
 			}
 
 		}

Bug#731967: libqb FTBFS: build failed on post-compile-test on mips/mipsel

[Dejan Latinovic]

Package: libqb
Version: 0.16.0.real-1
Severity: serious
Tags: sid patch
Justification: FTBFS


In an attempt to build libqb on mips/mipsel, 
build failed on testing:

 make  check-TESTS
 make[4]: Entering directory `/«PKGBUILDDIR»/tests'
 make[5]: Entering directory `/«PKGBUILDDIR»/tests'
 PASS: array.test
 PASS: map.test
 FAIL: rb.test
 PASS: log.test
 PASS: blackbox-segfault.sh
 PASS: loop.test
 PASS: ipc.test
 PASS: resources.test
 make[6]: Entering directory `/«PKGBUILDDIR»/tests'
 make[6]: Nothing to be done for `all'.
 make[6]: Leaving directory `/«PKGBUILDDIR»/tests'
 ==
 Testsuite summary for libqb 0.16.0
 ==
 # TOTAL: 8
 # PASS:  7
 # SKIP:  0
 # XFAIL: 0
 # FAIL:  1
 # XPASS: 0
 # ERROR: 0
 ==
 See tests/test-suite.log
 Please report to quarterback-de...@fedorahosted.org
 ==
 make[5]: *** [test-suite.log] Error 1

The full build logs are available from:
https://buildd.debian.org/status/fetch.php?pkg=libqbarch=mipsver=0.16.0.real-1stamp=1375430864
https://buildd.debian.org/status/fetch.php?pkg=libqbarch=mipselver=0.16.0.real-1stamp=1375430215


After I ran rb.test manually I got this error:

 Running suite(s): ringbuffer
 75%: Checks: 4, Failures: 0, Errors: 1
 check_rb.c:82:E:test01:test_ring_buffer1:0: (after this point) Received   
signal 11 (Segmentation fault)
 check_rb.c:123:P:test02:test_ring_buffer2:0: Passed
 check_rb.c:155:P:test03:test_ring_buffer3:0: Passed
 check_rb.c:183:P:test04:test_ring_buffer4:0: Passed


The problem is that test_ring_buffer1 in ./tests/check_rb.c,
at line:
 actual = qb_rb_chunk_read(rb, hdr, 512, 0);

is attempting to read chunk from ring buffer
 qb_ringbuffer_t *rb

and store it into address of struct hdr
 struct qb_ipc_request_header hdr;

If the size of the chunk(90-93) is larger than size of the hdr (16),
it comes to overwriting memory during reading.

This test does not fail on some other architectures
because there is a possibility that my_buf is located after hdr,
 char my_buf[512];
and then the overwriting passes unnoticed.


A patch fixing this issue is attached.


Regards,
Dejan Latinović
Author: Dejan Latinovic dejan.latino...@rt-rk.com
Description: Fix for ring buffer test.
Index: libqb-0.16.0.real/tests/check_rb.c
===
--- libqb-0.16.0.real.orig/tests/check_rb.c	2013-12-09 14:28:57.0 +
+++ libqb-0.16.0.real/tests/check_rb.c	2013-12-09 14:31:15.0 +
@@ -57,8 +57,9 @@
 hdr.id, actually the line number, i,
 __func__, __FILE__) + 1;
 			hdr.size += sizeof(struct qb_ipc_request_header);
+			memcpy(my_buf, hdr, sizeof(struct qb_ipc_request_header));
 			avail = qb_rb_space_free(rb);
-			actual = qb_rb_chunk_write(rb, hdr, hdr.size);
+			actual = qb_rb_chunk_write(rb, my_buf, hdr.size);
 			if (avail  (hdr.size + (3 * sizeof(uint32_t {
 ck_assert_int_eq(actual, -EAGAIN);
 			} else {
@@ -72,13 +73,13 @@
 		str = my_buf + sizeof(struct qb_ipc_request_header);
 
 		for (i = 0; i  15; i++) {
-			actual = qb_rb_chunk_read(rb, hdr, 512, 0);
+			actual = qb_rb_chunk_read(rb, my_buf, 512, 0);
 			if (actual  0) {
 ck_assert_int_eq(0, qb_rb_chunks_used(rb));
 break;
 			}
+			memcpy(hdr, my_buf, sizeof(struct qb_ipc_request_header));
 			str[actual - sizeof(struct qb_ipc_request_header)] = '\0';
-
 			ck_assert_int_eq(actual, hdr.size);
 		}
 	}

Bug#713743: lhapdf: FTBFS: octave/Makefile.am:9: error: using '$(srcdir)' in TESTS is currently broken: '$(srcdir)/lhapdf-octave-example1.m'

[Dejan Latinovic]

Control: tags -1 + patch 

Hi,

it also fails on other architectures with the same error.
I tried it on i386, mips and mipsel.

As I saw from build status for lhapdf, 
this package was buit successfully in the past :
https://buildd.debian.org/status/package.php?p=lhapdfsuite=sid

The problem is that version of automake was change in mean time,
from 1.11.6 to 1.14.

A patch fixing this issue is attached.


Also, tests are disabled for hurd-i386, mips, mipsel and s390
with:
DEB_BUILD_OPTIONS += nocheck
but is needed to add 
export DEB_BUILD_OPTIONS
into debian/rules file.


Regards,
Dejan LatinovićAuthor: Dejan Latinovic dejan.latino...@rt-rk.com
Description: Remove srcdir due to changes in automake.
diff -uNr lhapdf-5.8.9.orig/octave/Makefile.am lhapdf-5.8.9/octave/Makefile.am
--- lhapdf-5.8.9.orig/octave/Makefile.am	2013-12-03 15:17:38.0 +
+++ lhapdf-5.8.9/octave/Makefile.am	2013-12-03 15:18:48.0 +
@@ -6,4 +6,4 @@
 	$(MKOCTFILE) -I$(top_srcdir)/include -v -g -o lhapdf.oct wrapoctave.cc -L$(top_builddir)/lib -L$(top_builddir)/lib/.libs -lLHAPDF
 
 TESTS_ENVIRONMENT = LHAPATH=$(top_srcdir)/PDFsets LD_LIBRARY_PATH=$(top_builddir)/lib/.libs$${LD_LIBRARY_PATH:+:$$LD_LIBRARY_PATH} $(OCTAVE)
-TESTS = $(srcdir)/lhapdf-octave-example1.m
+TESTS = lhapdf-octave-example1.m

Bug#725631: getfem++: FTBFS due to: /usr/bin/install: cannot stat 'jar/*.jar': No such file or directory

[Dejan Latinovic]

Hi,

it seems that scilab does not work correctly,
here is snip of build log:

--
Making all in scilab
make[6]: Entering directory 
`/build/getfem++-uJr3zI/getfem++-4.2.1~beta1~svn4453~dfsg/interface/src/scilab'
SCI_DISABLE_TK=1 SCI_JAVA_ENABLE_HEADLESS=1 scilab -nw -nb -f 
/build/getfem++-uJr3zI/getfem++-4.2.1~beta1~svn4453~dfsg/interface/src/scilab/makefile_builder.sce
Error: Impossible to define SCIHOME environment variable.
 !--error 21 
Invalid index.
at line  72 of function listvarFunction called by :  
at line  29 of function listvarinfile called by :  
at line 826 of function %_sodload called by :  
at line   6 of function atomsSystemInit called by :  
  atomsSystemInit();
at line 107 of exec file called by :
exec('SCI/etc/scilab.start',-1);;exec('/build/getfem++-uJr3zI/getfem++-4.2.1~bet
 
 
 
 !--error 21 
Invalid index.
at line  72 of function listvarFunction called by :  
at line  29 of function listvarinfile called by :  
at line 826 of function %_sodload called by :  
at line   8 of function atomsQuit called by :  
at line   6 of function scilabQuit called by :  
scilabQuit();
at line  35 of exec file called by :
exec('SCI/etc/scilab.quit',-1);quit;
 
 
make[6]: Leaving directory 
`/build/getfem++-uJr3zI/getfem++-4.2.1~beta1~svn4453~dfsg/interface/src/scilab'
--


It looks like that scilab and scilab-include packages are excluded
from Build-Depends for mips architecture.
Could you do the same for mipsel?

Also it is needed to add mipsel in disable_scilab_archs list into debian rules 
file.

I attached all changes needed for mipsel.
With this changes, package was built successfully.


Could you consider applying these changes?



Regards,
Dejan Latinovićdiff -uNr debian.orig/control debian/control
--- debian.orig/control	2013-11-02 18:38:44.0 +
+++ debian/control	2013-12-04 15:47:55.0 +
@@ -22,8 +22,8 @@
  python-numpy,
  python-scipy,
  python-support (= 1.0.4),
- scilab [!mips !kfreebsd-amd64 !kfreebsd-i386 !hurd-i386 !s390x],
- scilab-include  [!mips !kfreebsd-amd64 !kfreebsd-i386 !hurd-i386 !s390x]
+ scilab [!mips !mipsel !kfreebsd-amd64 !kfreebsd-i386 !hurd-i386 !s390x],
+ scilab-include  [!mips !mipsel !kfreebsd-amd64 !kfreebsd-i386 !hurd-i386 !s390x]
 Standards-Version: 3.9.4
 Section: libs
 Homepage: http://home.gna.org/getfem/
@@ -116,7 +116,7 @@
 Pre-Depends:
  ${misc:Pre-Depends}
 Multi-Arch: same
-Architecture: alpha amd64 armel armhf i386 ia64 mips mipsel powerpc s390 s390x sparc
+Architecture: alpha amd64 armel armhf i386 ia64  powerpc s390 s390x sparc
 Depends:
  libgetfem4++ (= ${binary:Version}),
  scilab,
diff -uNr debian.orig/rules debian/rules
--- debian.orig/rules	2013-11-02 18:38:44.0 +
+++ debian/rules	2013-12-03 18:03:25.0 +
@@ -18,7 +18,7 @@
 		--enable-muparser --enable-mumps \
 		--with-mumps=-lsmumps_seq -ldmumps_seq -lcmumps_seq -lzmumps_seq -lpord_seq  
 
-disable_scilab_archs   = mips kfreebsd-amd64 kfreebsd-i386 s390x hurd-i386
+disable_scilab_archs   = mips mipsel kfreebsd-amd64 kfreebsd-i386 s390x hurd-i386
 ifneq (,$(filter $(DEB_HOST_ARCH), $(disable_scilab_archs)))
 	configure_flags += --disable-scilab
 else

Bug#712882: jaffl FTBFS:build failed on post-compile-test on mips

[Dejan Latinovic]

Control: tags -1 + patch jessie
Control: found -1 0.5.9-4

Hi,

Mips doesn't support neither -m64 nor -m32 options.

I have attached a patch solving the issue.

Regards,
Dejan Latinovićdiff -uNr jaffl-0.5.9.orig/libtest/GNUmakefile jaffl-0.5.9/libtest/GNUmakefile
--- jaffl-0.5.9.orig/libtest/GNUmakefile	2013-11-28 12:40:51.0 +
+++ jaffl-0.5.9/libtest/GNUmakefile	2013-11-26 16:44:45.0 +
@@ -121,10 +121,12 @@
 
 # On platforms (linux, solaris) that support both 32bit and 64bit, force building for one or the other
 ifneq ($(or $(findstring linux, $(OS)), $(findstring solaris, $(OS))),)
-  # Change the CC/LD instead of CFLAGS/LDFLAGS, incase other things in the flags
-  # makes the libffi build choke
-  CC += -m$(MODEL)
-  LD += -m$(MODEL)
+  ifneq ($(DEB_HOST_ARCH),$(filter $(DEB_HOST_ARCH),mips mipsel)) 
+# Change the CC/LD instead of CFLAGS/LDFLAGS, incase other things in the flags
+# makes the libffi build choke
+CC += -m$(MODEL)
+LD += -m$(MODEL)
+  endif
 endif
 
 LIBTEST = $(BUILD_DIR)/$(LIBNAME)

Bug#728869: gammaray FTBFS: build failed on post-compile-test on mips/mipsel

[Dejan Latinovic]

Package: gammaray
Version: 1.2.2-1
Severity: serious
Tags: jessie sid
Justification: FTBFS

In an attempt to rebuild the package on mips/mipsel, build failed on testing:

 Running tests...
 /usr/bin/ctest --force-new-ctest-process -j6
 Test project 
/build/buildd2-gammaray_1.2.2-1-mips-r9OCsw/gammaray-1.2.2/obj-mips-linux-gnu
 Start 1: connectiontest-preload
 Start 2: connectiontest-style
 Start 3: connectiontest-gdb
 Start 4: attachtest-gdb
 1/4 Test #1: connectiontest-preload ...***Failed2.43 sec
 Failed to launch injector preload
 Exit code: 1
 Error: Symbol is not marked as relocatable: qt_startup_hook
 
 2/4 Test #2: connectiontest-style .   Passed6.98 sec
 3/4 Test #3: connectiontest-gdb ...   Passed   13.83 sec
 4/4 Test #4: attachtest-gdb ...   Passed   33.19 sec
 
 75% tests passed, 1 tests failed out of 4
 
 Total Test time (real) =  33.30 sec
 
 The following tests FAILED:
 1 - connectiontest-preload (Failed)


The full build logs are available from:
https://buildd.debian.org/status/fetch.php?pkg=gammarayarch=mipsver=1.2.2-1stamp=1356200075
https://buildd.debian.org/status/fetch.php?pkg=gammarayarch=mipselver=1.2.2-1stamp=1356200317



GammaRay uses preloading to overwrite specific function calls in QtCore to
hook into its target application.

To be sure that the overwriting for qt_startup_hook will work,
test connectiontest-preload with readelf --relocs --wide checks if sybmbol
qt_startup_hook is marked as relocatable with JUMP_SLOT relocation, which
means that symbol will be resolved by dynamic linker and called at run
time through the plt.

The problem for Mips is that it has, besides the plt, another method of
calling functions from .so files, and this method doesn't need JUMP_SLOT
relocations (in fact, it doesn't need any relocations). This method uses .got
entries and lazy binding stubs. For example, if we call printf function, the
typical code is (gp register here points at the .got section):

lw t9, offset(gp)   // load in t9 function pointer from the address gp + offset
jalr t9 // call the function in t9

Initially, the memory at gp + offset will contain the address of the lazy
binding stub for printf. Lazy binding stubs look like:

lw  t9, 0(gp)
movet7,ra
li  t8,7
jalrt9

First instruction loads in t9 the address of the function _dl_runtime_resolve
from the predefined place in .got (it is located in first .got entry).
_dl_runtime_resolve is dynamic linker's function which resolves calls to
functions from .so files (printf in this case). Second and third instructions
load registers t7 and t8 with arguments that _dl_runtime_resolve needs. The key
argument here is 7 - it is dynamic symbol table index of printf, so that dynamic
linker knows which function to search for. Last instruction calls
_dl_runtime_resolve. After _dl_runtime_resolve finds printf in some .so file, it
will overwrite the .got entry for printf with found printf address, so that next
calls to printf will not go through lazy binding stub, but directly call printf.

All this doesn't need any relocations (below will be explained why). There is a
method how it can be determined whether a call to function will go through .got
and lazy binding stub.

First, Mips ABI specifies that .got is divided in two parts - local and global.
Dynamic linker only resolves symbols from the global part of the .got (that is,
only symbols from global part of the .got have lazy binding stubs. Entries from
the local part of the .got are filled with final values during static linking).
Mips ABI also specifies that the order of symbols in global part of the .got
must match the order of symbols in dynamic symbol table. In the .dynamic section
(obtained with readelf -d command), there is a Mips-specific dynamic tag
DT_MIPS_GOTSYM which gives the dynamic symbol index of the first symbol that has
entry in global part of the .got. So, the symbol with dynamic symbol index of
DT_MIPS_GOTSYM will be placed in the first entry in the global part of the .got.
The symbol with dynamic symbol index of DT_MIPS_GOTSYM+1 will be placed in the
second entry in the global part of the .got and so on.

For the previous example, lets suppose that DT_MIPS_GOTSYM is 6. Since the
dynamic symbol index of printf is 7, printf will be placed in the global part of
the .got (in the second entry of the global part of the got), and call to printf
will go through .got and lazy binding stub, and will be resolved by dynamic
linker.

From the previous explanation, it should be clear why calling functions through
.got on Mips doesn't need any relocations. Dynamic linker knows which .got entry
corresponds to which symbol, based on matching between .got and dynamic symbol
table. For the plt, on the other hand, there is no matching between the .got.plt
entry and the dynamic symbol table - function addresses in .got.plt can come in
arbitrary order. This is why every .got.plt entry 

Bug#724158: python-scrypt: FTBFS: scrypt-1.1.6/lib/crypto/crypto_aesctr.c:34:25: fatal error: openssl/aes.h: No such file or directory

[Dejan Latinovic]

Hi,
the same error occurs on mips/mipsel.

Full build log:
https://buildd.debian.org/status/fetch.php?pkg=python-scryptarch=mipsver=0.6.1-5stamp=1377211402

Header openssl/aes.h is part of libssl-dev package.
After I installed it, package was built successfully.

Is it a correct solution to add libssl-dev as a Build-Depends for
python-scrypt package?


Regards,
Dejan Latinović


--
To UNSUBSCRIBE, email to debian-bugs-rc-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org

Bug#665719: flint: FTBFS on mips/mipsel (impossible constraint in 'asm'

[Dejan Latinovic]

Hi,

gcc for mips/mipsel used in wheezy and jessie (sid) doesn't support
define_register_constraint h.


I have attached a patch solving the issue.

Cheers,
Dejan Latinovićdiff -uNr flint-1.011.orig/longlong.h flint-1.011/longlong.h
--- flint-1.011.orig/longlong.h	2008-06-16 03:31:56.0 +
+++ flint-1.011/longlong.h	2013-10-23 15:48:54.0 +
@@ -1011,27 +1011,17 @@
 #endif /* __m88000__ */
 
 #if defined (__mips)  W_TYPE_SIZE == 32
-#if __GNUC__  2 || __GNUC_MINOR__ = 7
-#define umul_ppmm(w1, w0, u, v) \
-  __asm__ (multu %2,%3 : =l (w0), =h (w1) : d (u), d (v))
-#else
 #define umul_ppmm(w1, w0, u, v) \
   __asm__ (multu %2,%3\n\tmflo %0\n\tmfhi %1\
 	   : =d (w0), =d (w1) : d (u), d (v))
-#endif
 #define UMUL_TIME 10
 #define UDIV_TIME 100
 #endif /* __mips */
 
 #if (defined (__mips)  __mips = 3)  W_TYPE_SIZE == 64
-#if __GNUC__  2 || __GNUC_MINOR__ = 7
-#define umul_ppmm(w1, w0, u, v) \
-  __asm__ (dmultu %2,%3 : =l (w0), =h (w1) : d (u), d (v))
-#else
 #define umul_ppmm(w1, w0, u, v) \
   __asm__ (dmultu %2,%3\n\tmflo %0\n\tmfhi %1\
 	   : =d (w0), =d (w1) : d (u), d (v))
-#endif
 #define UMUL_TIME 20
 #define UDIV_TIME 140
 #endif /* __mips */

Bug#727657: unconditional build dependency on g++-multilib

[Dejan Latinovic]

Hi,

  uc-echo has an unconditional build dependency on g++-multilib, which
is not
  available on all architectures.
 do you know a better choice?


maybe the better way would be to list all architectures that support
g++-multilib.
For sid those architectures are:  amd64, i386, kfreebsd-amd64, mips,
mipsel, powerpc, ppc64, s390x, sparc, sparc64, x32
For example, g++-multilib [amd64 i386 kfreebsd-amd64 mips mipsel powerpc
ppc64 s390x sparc sparc64 x32]


  also, it fails to build on mips* which does have g++-multilib, so it
seems that
  this package has hardcoded ix86 assumptions.

 The problem on mips* is not g++-multilib but the parameter -m64 for gcc.
 Somewhere I read that -mabi=64 is default on mips, but 'gcc -dumpspecs'
 tells me:
  *multilib_defaults:
  EB mips1 mabi=32
 I guess I am totally confused now :-(. Can you please point me in the
 right direction?

Mips doesn't support neither -m64 nor -m32 options.
For MIPS64 you should use -mips64 -mabi=64.
For MIPS32 you should use defaults options.


Cheers,
Dejan Latinović


--
To UNSUBSCRIBE, email to debian-bugs-rc-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org