Bug#1077690: gtk4: intended CFLAGS/LDFLAGS no longer set, leading to test failures and FTBFS on i386

[Simon McVittie]

Source: gtk4
Version: 4.14.4-5
Severity: serious
Tags: ftbfs
Justification: fails to build from source (but built successfully in the past)
X-Debbugs-Cc: Adrian Bunk <bunk at debian.org>

(Redirecting from a reply to #1077675 since this is a different bug)

On Thu, 01 Aug 2024 at 00:19:45 +0300, Adrian Bunk wrote:
> On Wed, Jul 31, 2024 at 08:56:55PM +0100, Simon McVittie wrote:
> >,,,
> > The good news is that this seems to be the only test failure, on all the
> > architectures that have been attempted so far.
> 
> The bad news is that this is not true on i386,
> apparently the -ffloat-store is gone there.
> 
> Comparing build logs the "-Wl,-z,defs -Wl,-O1" is also gone
> (not only on i386), so I'd suspect some of the recent dpkg
> changes in that area might have broken LDFLAGS/CFLAGS setting?

I think this must be the case. I built 4.14.4-4 successfully for i386 while
testing the -ffloat-store change, so it must be a very recent change
(since the weekend) that made it regress.

Perhaps debian/rules will need some reordering so that it gets the
architectures from architecture.mk, then uses those to find out whether
we are on i386, then sets DEB_CFLAGS_MAINT_APPEND (on i386) and
DEB_LDFLAGS_MAINT_APPEND (everywhere), and finally includes the rest of
default.mk so the build flags are set up?

If a dpkg-dev change can make our intended CFLAGS and LDFLAGS go from
"works fine" to "silently ignored", that feels uncomfortably fragile.

    smcv