On 2010/6/3 Adam C Powell IV wrote:
[...]
>> A patch will look like
>> --- debian/rules
>> +++ debian/rules
>> @@ -84,7 +84,10 @@ clean:
>>         rm -f *-stamp
>>         dh_clean
>>
>> -configure-stamp:
>> +patch-stamp:
>> +       QUILT_PATCHES=debian/patches quilt push -a || test $$? = 2
>> +
>> +configure-stamp: patch-stamp
>>         dh_testdir
>>  #      Move aside obsolete files
>>         for obsoletefile in KERNEL_SRC_$(SALOME_VERSION)/bin/runSalome \
>>
>> But there are other problems, I will try to provide working patches tonight.
>
> Wonderful, thanks!

Here are several patches:
 * 0001 is the patch mentioned above
 * 0002 and 0003 make build abort when something goes wrong
 * 0004 implements what is described in
/usr/share/doc/autotools-dev/README.Debian.gz as best practice.  It is
more complicated than I would like, because of the way upstream uses
autotools.  It has not been fully tested.
 * 0005 tries to make removal of generated files simpler, it has been
even less tested.
Thus I suggest you to commit 0001-003.  0004 could also be committed
after review.  About 0005, you decide whether building out of source
tree is a good idea.

This will not fix all FTBFS, I also saw problems related to doxygen.
I did not have time to investigate yet, but I believe that docs are
also generated when running the build-arch target; doxygen will not be
installed on autobuilders and build will fail.

Denis

Attachment: 0001-Fix-FTBFS-patches-must-be-applied-before-building.patch
Description: Binary data

Attachment: 0002-Abort-in-clean-target-if-quilt-pop-fails.patch
Description: Binary data

Attachment: 0003-Abort-when-a-module-cannot-be-configured.patch
Description: Binary data

Attachment: 0004-DO-NOT-COMMIT-NOT-FULLY-TESTED-In-clean-target-remov.patch
Description: Binary data

Attachment: 0005-DO-NOT-COMMIT-NOT-FULLY-TESTED-Build-out-of-tree.patch
Description: Binary data

Reply via email to