Bug#974793: [Pkg-opencl-devel] Bug#974793: Bug#974793: oclgrind: Please upgrade to llvm-toolchain-11

[James Price]

I've started adding LLVM 11 support upstream today. I will try and get this
finished soon and then push out a new upstream release that we can package
here.

James


On Sun., Nov. 15, 2020, 19:27 Andreas Beckmann,  wrote:

> Control: tag -1 upstream
>
> oclgrind FTBFS with both LLVM 10 and LLVM 11.
>
> First errors:
>
> In file included from /usr/lib/llvm-10/include/llvm/ADT/StringRef.h:12,
>  from
> /usr/lib/llvm-10/include/llvm/IR/DiagnosticHandler.h:15,
>  from /usr/lib/llvm-10/include/llvm/IR/LLVMContext.h:18,
>  from /build/oclgrind-19.10/src/core/Context.cpp:21:
> /usr/lib/llvm-10/include/llvm/ADT/STLExtras.h:559:49: error:
> 'std::index_sequence' has not been declared
>   559 |   template  value_type
> deref(std::index_sequence) const {
>   | ^~
>
> In file included from /usr/lib/llvm-11/include/llvm/ADT/Hashing.h:49,
>  from /usr/lib/llvm-11/include/llvm/ADT/ArrayRef.h:12,
>  from /usr/lib/llvm-11/include/llvm/IR/Instructions.h:18,
>  from /build/oclgrind-19.10/src/core/common.cpp:18:
> /usr/lib/llvm-11/include/llvm/Support/SwapByteOrder.h:152:13: error:
> 'enable_if_t' in namespace 'std' does not name a template type
>   152 | inline std::enable_if_t::value, T>
> getSwappedBytes(T C) {
>   | ^~~
> /usr/lib/llvm-11/include/llvm/Support/SwapByteOrder.h:152:8: note:
> 'std::enable_if_t' is only available from C++14 onwards
>   152 | inline std::enable_if_t::value, T>
> getSwappedBytes(T C) {
>   |^~~
>
>
> Andreas
>
> ___
> Pkg-opencl-devel mailing list
> pkg-opencl-de...@alioth-lists.debian.net
> https://alioth-lists.debian.net/cgi-bin/mailman/listinfo/pkg-opencl-devel

Bug#882875: [Pkg-opencl-devel] Bug#882875: pocl: FTBFS with cmake: kernel/test_printf fails on i386

[James Price]

Reproduced on my 32-bit VM.

The issue is that the top level CMakeLists.txt checks CMAKE_SYSTEM_PROCESSOR to 
determine whether the machine is i386, but on some systems the processor is 
actually i686 or similar. Probably just needs to match against "i.86” instead, 
which is what the old `fix-tests-i386.patch` used to do for autotools.

James

> On 27 Nov 2017, at 18:31, Andreas Beckmann  wrote:
> 
> Source: pocl
> Version: 0.13-12
> Severity: serious
> Justification: fails to build from source (but built successfully in the past)
> 
> https://buildd.debian.org/status/fetch.php?pkg=pocl=i386=0.13-12=1511803243=0
> 
> Can't reproduce this on my amd64 machine in a i386 chroot.
> 
> It may be related to set-POCL_DEVICE_ADDRESS_BITS-before-use.patch which
> fixed that for me.
> 
> As I understood, the test is supposed to fail on i386 and succeed on
> amd64. That is being fixed up by fix-tests-i386.patch for autotest, and
> tests/kernel/CMakeLists.txt also should set WILL_FAIL 1 as well.
> Unless I386 detection is still not working.
> 
> Packaging is in branch experimental.
> 
> 
> Andreas
> 
> PS: I want this to work like 0.13-11 in sid before I even start
>debugging new issues with 0.14
> 
> ___
> Pkg-opencl-devel mailing list
> pkg-opencl-de...@lists.alioth.debian.org
> http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/pkg-opencl-devel

Bug#873417: [Pkg-opencl-devel] Bug#873417: pocl: Please update to llvm-toolchain 4.0 or, better, 5.0

[James Price]

The main issue seems to be that pocl has switched to CMake. I’ve fixed this 
locally and it builds and passes all the tests, it just looks like the symbols 
files need to be updated which I’m not sure how to do properly.

I don’t have write access to collab-maint so couldn’t push the CMake fix, so 
patch is attached if anyone is interested.

James



> On 21 Nov 2017, at 07:59, Andreas Tille  wrote:
>
> Hi,
>
> I realised that there is a new upstream version 0.14 which seems to
> allow LLVM 4.0.  I gave it a try in Git in the new branch 0.14 but faced
> build problems I was not able to solve.  However, since I was hoping
> that the refreshed patches might help (please review!) I commited this
> branch.
>
> Hope this helps a bit
>
>  Andreas.
>
> --
> http://fam-tille.de
>
> ___
> Pkg-opencl-devel mailing list
> pkg-opencl-de...@lists.alioth.debian.org
> http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/pkg-opencl-devel



pocl-cmake.diff
Description: pocl-cmake.diff

Bug#807593: RM: oclgrind [mips powerpc s390x] -- ROM; broken binaries

[James Price]

Package: ftp.debian.org
Severity: normal

Please remove out-of-date binaries for oclgrind 15.5-1 for mips,
powerpc and s390x as they are broken and preventing migration of a
newer version.

The 15.5-1 version was missing a build dependency which meant that
testing was disabled, and therefore it didn't FTBFS on these
architectures when it really should have.

Bug#806965: oclgrind: FTBFS on ppc64el -- conflict with altivec keyword bool

[James Price]

Thanks for the report. I’ve actually applied a slightly different fix for this 
a little while back (see [1]), but that revision has yet to be uploaded.

The fix I applied was essentially the same as has been done for the 
opencl-headers package, described in [2]. This just uses __vector instead of 
vector to avoid conflicts with similarly named types.

If you think the '-mno-altivec’ fix is preferable then I can revert my patch 
and apply yours instead. Otherwise I’ll just request that the latest revision 
be uploaded.

Many thanks,

James

[1] 
https://anonscm.debian.org/cgit/pkg-opencl/oclgrind.git/commit/?id=cd9925f14b76d3867166b524b3305ae156f5ef6b
 

[2] https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=760310 

Bug#799322: pocl: With 0.11 some test are failing

[James Price]

tag 799322 patch
thanks

On Wed, 14 Oct 2015 18:33:01 +0200 Gert Wollny  wrote:
> Hi, I had a look into the package, but I for me with 0.11 an llvm-3.6
> two tests are failing - this has also already been reported upstream:
> 
> https://github.com/pocl/pocl/issues/181
> 

I’ve attached a patch that fixes these test failures. With this patch, and 
removing the libpocl1.symbols file as mentioned earlier, I can now build the 
0.11 package successfully. I’ve forwarded the patch upstream.

There is now a 0.12 release as well, so I’m not sure whether you intend to 
finish the 0.11 package first or just skip straight to 0.12. The 0.12 release 
seems to build cleanly for me and passes all the tests (with the same patch 
applied).

James




fix-image-args.patch
Description: Binary data

Bug#794905: llvm-3.7-dev: Incorrect install prefix in installed LLVMConfig.cmake

[James Price]

Package: llvm-3.7-dev
Version: 1:3.7~+rc2-2
Severity: normal
Tags: patch

Dear Maintainer,

The LLVM_INSTALL_PREFIX variable set by LLVMConfig.cmake wrongly
assumes that the installed CMake configuration files are in a
subdirectory of the actual LLVM installation. The Debian package moves
the CMake configuration files to /usr/share/llvm-VERSION/, so this
assumption is not correct.

The way that the LLVM_INSTALL_PREFIX variable is determined has
changed in LLVM 3.7, and this issue for Debian was pointed out (along
with a proposed fix) by Brad King, here:

http://thread.gmane.org/gmane.comp.compilers.llvm.cvs/255502/focus=25574

The attached patch resolves this by essentially using the same
mechanism to determine LLVM_INSTALL_PREFIX as was used in LLVM
3.6. The resulting changes to the generated LLVMConfig.cmake match the
changes proposed by Brad above.

Many thanks,

James

-- System Information:
Debian Release: stretch/sid
  APT prefers unstable
  APT policy: (500, 'unstable')
Architecture: amd64 (x86_64)

Kernel: Linux 4.1.0-1-amd64 (SMP w/8 CPU cores)
Locale: LANG=en_GB.UTF-8, LC_CTYPE=en_GB.UTF-8 (charmap=UTF-8)
Shell: /bin/sh linked to /bin/dash
Init: systemd (via /run/systemd/system)

Versions of packages llvm-3.7-dev depends on:
ii  libc6  2.19-19
ii  libedit2   3.1-20150325-1
ii  libffi-dev 3.2.1-3
ii  libffi63.2.1-3
ii  libgcc11:5.2.1-14
ii  libjsoncpp0v5  0.10.5-1
ii  libllvm3.7 1:3.7~+rc2-2
ii  libstdc++6 5.2.1-14
ii  libtinfo-dev   5.9+20150516-2
ii  libtinfo5  5.9+20150516-2
ii  llvm-3.7   1:3.7~+rc2-2
ii  zlib1g 1:1.2.8.dfsg-2+b1

llvm-3.7-dev recommends no packages.

llvm-3.7-dev suggests no packages.

-- no debconf information
--- a/cmake/modules/Makefile
+++ b/cmake/modules/Makefile
@@ -62,17 +62,9 @@
 
 LLVM_CONFIG_CODE := \
 \# Compute the CMake directory from the LLVMConfig.cmake file location.\n\
-get_filename_component(_LLVM_CMAKE_DIR $${CMAKE_CURRENT_LIST_FILE} PATH)\n\
-\# Compute the installation prefix from the LLVMConfig.cmake file location.\n\
-get_filename_component(LLVM_INSTALL_PREFIX $${CMAKE_CURRENT_LIST_FILE} PATH)\n
+get_filename_component(_LLVM_CMAKE_DIR $${CMAKE_CURRENT_LIST_FILE} PATH)\n
 
-# Compute number of levels (typically 3 - ``share/llvm/cmake/``) to PROJ_prefix
-# from PROJ_cmake, then emit the appropriate number of calls to
-# get_filename_components(). Note this assumes there are no spaces in the
-# cmake_path_suffix variable.
-cmake_path_suffix := $(subst $(PROJ_prefix),,$(subst $(DESTDIR),,$(PROJ_cmake)))
-cmake_path_dirs := $(subst /, ,$(cmake_path_suffix))
-LLVM_CONFIG_CODE += $(foreach __not_used,$(cmake_path_dirs),get_filename_component(LLVM_INSTALL_PREFIX $${LLVM_INSTALL_PREFIX} PATH)\n)
+LLVM_CONFIG_CODE += set(LLVM_INSTALL_PREFIX $(subst /,\/,$(PROJ_prefix)))\n
 
 LLVM_CONFIG_CODE += set(_LLVM_LIBRARY_DIR $${LLVM_INSTALL_PREFIX}\/lib)
 

Bug#794787: ITP: oclgrind -- OpenCL device simulator

[James Price]

Package: wnpp
Severity: wishlist
Owner: James Price j.pr...@bristol.ac.uk

* Package name: oclgrind
  Version : 15.5
  Upstream Author : James Price j.pr...@bristol.ac.uk
* URL : https://github.com/jrprice/Oclgrind
* License : BSD
  Programming Lang: C++
  Description : OpenCL device simulator

Oclgrind is an extensible OpenCL device simulator that provides a
plugin interface to facilitate the creation of tools to aid analysis
and development of OpenCL programs. Among the tools that Oclgrind
provides are various debugging aids, such as out-of-bounds memory
access checking, data-race detection, and an interactive debugger.


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