[Buildd-tools-devel] Bug#482491: Bug#482491: Bug#482491: sbuild: extracts the source while installing b-deps

[Roger Leigh]

Lucas Nussbaum <lucas at lucas-nussbaum.net> writes:

> On 23/05/08 at 12:08 +0100, Roger Leigh wrote:
>> On Fri, May 23, 2008 at 09:25:01AM +0200, Lucas Nussbaum wrote:
>> > It seems that sbuild now extracts the source while installing
>> > build-deps. See:
>> > 
>> > Selecting previously deselected package gettext.
>> > Unpacking gettext (from .../gettext_0.17-2_i386.deb) ...
>> > Selecting previously deselected package intltool-debian.
>> > Unpacking intltool-debian (from .../intltool-debian_0.35.0+20060710gpg:
>> > Signature made Wed Apr 19 12:53:15 2006 CEST using DSA key ID 8E635A5E
>> > gpg: Can't check signature: public key not found
>> > dpkg-source: extracting crip in crip_3.7-3
>> > dpkg-source: info: unpacking crip_3.7.orig.tar.gz
>> > dpkg-source: info: applying crip_3.7-3.diff.gz
>> > .1_all.deb) ...
>> > Selecting previously deselected package po-debconf.
>> > Unpacking po-debconf (from .../po-debconf_1.0.13_all.deb) ...
>> > Selecting previously deselected package debhelper.
>> > Unpacking debhelper (from .../debhelper_7.0.9_all.deb) ...
>> > 
>> > This produces confusing build log. If this change was made on purpose,
>> > please provide an option to serialize those steps, so I can get "clean"
>> > build logs when reporting bugs.
>> 
>> There have not been any changes made on purpose.  install_deps() is
>> called before build().  One possiblity is that the logging stream PLOG
>> is doing some odd buffering in subprocesses, and flushing the stream at
>> certain points would help (the logging is done in a separate process).
>> 
>> It would be interesting to know the the git master head 
>>   git://git.debian.org/git/buildd-tools/sbuild
>> also suffers from this problem, and if the log is identical between
>> sbuild runs.
>> 
>> sbuild has had quite a lot of refactoring done over the last two weeks
>> or so, and it's possible that something broke as a result, but none of
>> this work should have changed the program logic.
>
> git bisect shows that the first bad commit is
> c1e3cbc8699da54f9f540bdabc6436a128eaa668 .
>
> A test case is to build crip_3.7-3 : it builds fine before the commit,
> and fails to build with the commit.

Is this for #482456 rather than this bug?

For me, crip fails to apply a dpatch patch (log attached) using the
current git (master).

However, I do see dpkg-source logged as running /during/
dpkg-buildpackage.  I still suspect this is a buffering issue in the
PLOG stream, but I don't have any remaining time this weekend to track
it down, I'm afraid.  This certainly shouldn't cause build failures
though, just out-of-order messages in the logfile.


Regards,
Roger

-- 
  .''`.  Roger Leigh
 : :' :  Debian GNU/Linux             http://people.debian.org/~rleigh/
 `. `'   Printing on GNU/Linux?       http://gutenprint.sourceforge.net/
   `-    GPG Public Key: 0x25BFB848   Please GPG sign your mail.
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: crip_3.7-3_20080525-2305
Url: http://lists.alioth.debian.org/pipermail/buildd-tools-devel/attachments/20080525/aee4babf/attachment.txt 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 188 bytes
Desc: not available
Url : http://lists.alioth.debian.org/pipermail/buildd-tools-devel/attachments/20080525/aee4babf/attachment.pgp