Bug#805193: aster: FTBFS: Fatal Error: finclude/petscsys.h: No such file or directory

[Graham Inggs]

Control: tags -1 pending

I've committed the fix to git and will upload once libopenmpi1.10 is 
available (see #813042).

Bug#805193: aster: FTBFS: Fatal Error: finclude/petscsys.h: No such file or directory

[Graham Inggs]

Hi Peter

On 13 December 2015 at 01:17, peter green  wrote:
> On 12/12/15 19:01, peter green wrote:
>>
>>
>> New debdiff attatched, still no intent to NMU.
>
> Sorry, debdiff at
> http://debdiffs.raspbian.org/main/a/aster/aster_11.5.0%2bdfsg2-3%2brpi1.debdiff

Thanks for your work on this!

If there are no objections, I'll do a team upload in a couple of days.

Regards
Graham

Bug#805193: aster: FTBFS: Fatal Error: finclude/petscsys.h: No such file or directory

[peter green]

On 12/12/15 17:32, peter green wrote:
I was able to get aster to build for raspbian stretch (in which I have 
imported the updated petsc from debian sid) with the changes in the 
attatched debdiff.
Sorry I must learn to read the screen more carefully and actually check 
that something has succeeded. The build in fact seems to have failed 
with an "unrepresentable changes to source" error. Further the debdiff I 
sent you doesn't represent the state of my source tree at the time I 
sent it due to said error.


There was another fix to the clean target needed. I am now testing the 
build again and will report back later with whether it suceeded or not.

Bug#805193: aster: FTBFS: Fatal Error: finclude/petscsys.h: No such file or directory

[peter green]

Tags 805193 +patch
Thanks

I was able to get aster to build for raspbian stretch (in which I have 
imported the updated petsc from debian sid) with the changes in the 
attatched debdiff. I have not tried it on any other system. I have not 
tested if the resulting packages actually work. I will be uploading this 
to raspbian shortly. No intent to NMU in Debian.


The failure in the bug report was fixed by adding an include path. I 
added the include path to three different variables. I'm not sure which 
of them the build is actually using.


I then ran into another failure, PETSC_DEFAULT_DOUBLE_PRECISION seems to 
have dissapeared. I did some googling and found 
http://lists.mcs.anl.gov/pipermail/petsc-users/2014-February/020567.html 
which seemed to indicate that it should be replaced with 
PETSC_DEFAULT_REAL. I did so and got a successful build.


I also ran into and fixed a broken clean target while working on the 
package. I forgot to document this in the changelog.


P.S. Trying to build this package as root will break your system's 
python-numpy and then fail because of said breakage.
diff -Nru aster-11.5.0+dfsg2/debian/changelog 
aster-11.5.0+dfsg2/debian/changelog
--- aster-11.5.0+dfsg2/debian/changelog 2014-08-04 10:18:27.0 +
+++ aster-11.5.0+dfsg2/debian/changelog 2015-12-11 03:19:31.0 +
@@ -1,3 +1,10 @@
+aster (11.5.0+dfsg2-3+rpi1) stretch-staging; urgency=medium
+
+  * Force new include path for petsc fortran stuff.
+  * Change PETSC_DEFAULT_DOUBLE_PRECISION to PETSC_DEFAULT_REAL as hinted at 
by http://lists.mcs.anl.gov/pipermail/petsc-users/2014-February/020567.html
+
+ -- Peter Michael Green   Fri, 11 Dec 2015 00:53:43 
+
+
 aster (11.5.0+dfsg2-3) unstable; urgency=medium
 
   * Team upload.
diff -Nru aster-11.5.0+dfsg2/debian/patches/0011-petsc-default.patch 
aster-11.5.0+dfsg2/debian/patches/0011-petsc-default.patch
--- aster-11.5.0+dfsg2/debian/patches/0011-petsc-default.patch  1970-01-01 
00:00:00.0 +
+++ aster-11.5.0+dfsg2/debian/patches/0011-petsc-default.patch  2015-12-11 
03:35:46.0 +
@@ -0,0 +1,16 @@
+Description:  Change PETSC_DEFAULT_DOUBLE_PRECISION to PETSC_DEFAULT_REAL as 
hinted at by 
http://lists.mcs.anl.gov/pipermail/petsc-users/2014-February/020567.html
+Author: Peter Michael Green 
+
+--- aster-11.5.0+dfsg2.orig/bibfor/petsc/apksp.F90
 aster-11.5.0+dfsg2/bibfor/petsc/apksp.F90
+@@ -103,8 +103,8 @@ subroutine apksp(kptsc)
+ endif
+ !
+ rtol = resire
+-atol = PETSC_DEFAULT_DOUBLE_PRECISION
+-dtol = PETSC_DEFAULT_DOUBLE_PRECISION
++atol = PETSC_DEFAULT_REAL
++dtol = PETSC_DEFAULT_REAL
+ !
+ call KSPSetTolerances(ksp, rtol, atol, dtol, maxits,&
+   ierr)
diff -Nru aster-11.5.0+dfsg2/debian/patches/series 
aster-11.5.0+dfsg2/debian/patches/series
--- aster-11.5.0+dfsg2/debian/patches/series2014-08-04 10:07:55.0 
+
+++ aster-11.5.0+dfsg2/debian/patches/series2015-12-11 03:35:26.0 
+
@@ -8,3 +8,4 @@
 0008-Disable-BLACS-checks.patch
 0009-Fix-NumPy-version-comparison.patch
 0010-Check-that-DEST_CPU-exists-before-reading-it.patch
+0011-petsc-default.patch
diff -Nru aster-11.5.0+dfsg2/debian/rules aster-11.5.0+dfsg2/debian/rules
--- aster-11.5.0+dfsg2/debian/rules 2014-08-04 10:13:49.0 +
+++ aster-11.5.0+dfsg2/debian/rules 2015-12-11 03:01:43.0 +
@@ -11,6 +11,9 @@
 
 ifneq ($(wildcard 
/usr/lib/$(DEB_HOST_MULTIARCH)/hdf5/$(ARCH_DEFAULT_MPI_IMPL)/libhdf5.so),)
   export DEB_CPPFLAGS_MAINT_APPEND := 
-I/usr/include/hdf5/$(ARCH_DEFAULT_MPI_IMPL)
+  export DEB_FCFLAGS_MAINT_APPEND :=  
-I/usr/lib/petscdir/3.6.2/$(DEB_HOST_MULTIARCH)-real/include/petsc/
+  export DEB_FFLAGS_MAINT_APPEND :=  
-I/usr/lib/petscdir/3.6.2/$(DEB_HOST_MULTIARCH)-real/include/petsc/
+  export DEB_CFLAGS_MAINT_APPEND :=  
-I/usr/lib/petscdir/3.6.2/$(DEB_HOST_MULTIARCH)-real/include/petsc/
   export DEB_LDFLAGS_MAINT_APPEND := 
-Wl,-L/usr/lib/$(DEB_HOST_MULTIARCH)/hdf5/$(ARCH_DEFAULT_MPI_IMPL)
 endif
 

Bug#805193: aster: FTBFS: Fatal Error: finclude/petscsys.h: No such file or directory

[peter green]

On 12/12/15 17:47, peter green wrote:

On 12/12/15 17:32, peter green wrote:
I was able to get aster to build for raspbian stretch (in which I 
have imported the updated petsc from debian sid) with the changes in 
the attatched debdiff.
Sorry I must learn to read the screen more carefully and actually 
check that something has succeeded. The build in fact seems to have 
failed with an "unrepresentable changes to source" error. Further the 
debdiff I sent you doesn't represent the state of my source tree at 
the time I sent it due to said error.


There was another fix to the clean target needed. I am now testing the 
build again and will report back later with whether it suceeded or not.
Another gotcha I ran into, for some reason even after I su'd to nobody 
it still wanted to use the same temp directory (with root in it's name) 
which already existed and was owned by root causing the build to fail.


So I removed that directory from /tmp and tried the build again and 
FINALLY got a successful build.


New debdiff attatched, still no intent to NMU.

Bug#805193: aster: FTBFS: Fatal Error: finclude/petscsys.h: No such file or directory

[peter green]

On 12/12/15 19:01, peter green wrote:


New debdiff attatched, still no intent to NMU.
Sorry, debdiff at 
http://debdiffs.raspbian.org/main/a/aster/aster_11.5.0%2bdfsg2-3%2brpi1.debdiff