After an evaluation, GNOME has moved from Bugzilla to GitLab. Learn more about GitLab.
No new issues can be reported in GNOME Bugzilla anymore.
To report an issue in a GNOME project, go to GNOME GitLab.
Do not go to GNOME Gitlab for: Bluefish, Doxygen, GnuCash, GStreamer, java-gnome, LDTP, NetworkManager, Tomboy.
Bug 327515 - distribution broken for builds where builddir != srcdir
distribution broken for builds where builddir != srcdir
Status: RESOLVED FIXED
Product: libxslt
Classification: Platform
Component: general
1.1.15
Other Linux
: Normal normal
: ---
Assigned To: Daniel Veillard
libxml QA maintainers
Depends on:
Blocks: 414139
 
 
Reported: 2006-01-18 11:24 UTC by tim.vanholder
Modified: 2014-08-26 10:18 UTC
See Also:
GNOME target: ---
GNOME version: ---



Description tim.vanholder 2006-01-18 11:24:33 UTC
The tarball for libxslt 1.1.15 ships with libxslt/xsltconfig.h (and
libxslt/xsltwin32config.h) included.  These are normally generated by configure,
so should not be included in a distribution.  In particular, they override the
files created by configure if builddir != distdir (leading to build failure
if libxml2 has no module support, because the shipped headers use have
WITH_MODULES = 1).
Comment 1 tim.vanholder 2006-01-18 11:25:51 UTC
typo: should be "builddir != srcdir", not "builddir != distdir" in the above comment
Comment 2 Daniel Veillard 2006-01-18 15:34:51 UTC
xsltwin32config.h  is needed for MSC windows users, who can't use configure.
Not a bug !
Not GNOME either !
Certainly not a major bug from a GNOME point of view.

90+% of my windows users are using MSC to compile their code I guess.
If you start by doing crazy stuff like cross-compiling for windows on
Linux, then the very first thing to do is to not make things even more
twisted by not building in the source dir, nuff' said !

Daniel
Comment 3 tim.vanholder 2006-01-18 18:36:20 UTC
If you read the original report, you will see that my main mention is of
libxslt/xsltconfig.h, with xsltwin32config.h only mentioned as an aside (and
for that file I certainly accept the "win32 users can't run configure" argument).

I also made no mention of cross-compilation as it affected both a straight
build and a cross build.  The header hides the one that is generated by
configure, which is most certainly a bug, and would affect anyone who specifies
specific configure options (such as --without-plugins) while building outside
$srcdir (which is quite a reasonable situation, especially if a package needs
to be built in multiple configurations (gcc3, gcc4, cross, ...)).

If you feel that anything non-cookie-cutter is not Gnome, then fine, I will
resubmit on the mailing list - but this IS a bug.
Comment 4 Daniel Veillard 2006-01-18 20:25:50 UTC
Okay then this need to be reopened, I think I undestand the issue better now,
can probably be done easilly by removing the source version within configure.in
before regenerating them

Daniel
Comment 5 tim.vanholder 2006-01-19 07:04:51 UTC
Well that would work, except that it would break for a read-only srcdir.

Are both headers only in the distribution for non-configure Windows targets?
If so, why not put them in the win32 subdir and have configure.js copy them in
place (or use -I$srcdir/win32 to pull them in)?  They'd still be available that
way, without interfering with "normal" configure-based builds.
Comment 6 Daniel Veillard 2006-01-19 07:46:52 UTC
> Are both headers only in the distribution for non-configure Windows targets?
yes

But anything changing the way Windows build or use works should really be
discussed on the list, I can't represent those users viewpoint !

Daniel
Comment 7 tim.vanholder 2006-01-19 08:30:28 UTC
> > Are both headers only in the distribution for non-configure Windows targets?
> yes
> 
> But anything changing the way Windows build or use works should really be
> discussed on the list, I can't represent those users viewpoint !

OK - I'll send a report of this problem to the list; if/when a solution is
agreed there (and implemented), this item can then be closed.
Comment 8 tim.vanholder 2014-08-26 10:18:13 UTC
Problem seems to be gone in most recent version 1.1.28 (Nov 2012), so marking as fixed.