On 4/6/24 1:29 PM, Aurelien Jarno wrote:
On 2024-04-06 08:01, Sebastiaan Couwenberg wrote:
On 4/5/24 9:51 PM, Aurelien Jarno wrote:
For Bookworm given we can not fix the compiler easily, I propose to just
build icinga2 with -O1 on ppc64el. If you are fine with that option, I
can take care of proposing a patch and submitting it to the stable
release team.
A patch for this is very welcome. How do you propose to implement that?
Something like this maybe?
--- a/debian/rules
+++ b/debian/rules
@@ -9,6 +9,11 @@ include /usr/share/dpkg/architecture.mk
export CTEST_OUTPUT_ON_FAILURE=1
+ifneq (,$(filter $(DEB_HOST_ARCH), ppc64el))
+ export DEB_CXXFLAGS_STRIP = -O2
+ export DEB_CXXFLAGS_MAINT_APPEND = -O1
+endif
+
ifneq (,$(filter $(DEB_HOST_ARCH), armel mips mipsel powerpc))
export DEB_LDFLAGS_MAINT_APPEND += -Wl,--no-as-needed -latomic
-Wl,--as-needed
endif
Yes, something like that works. I even tested without the
DEB_CXXFLAGS_STRIP, gcc is smart enough to just take the last flag, so
-O1.
Also it seems that your diff applies to the Trixie/Sid version, while it
should be applied to Bookworm instead.
Correct, I did not actually prepare a bookworm branch as you offered to
take care of that.
Since it's not that much work, I did that now:
https://salsa.debian.org/nagios-team/icinga2/-/commits/bookworm?ref_type=heads
If you can confirm that those changes fix the issue, I can also fix the
bookworm-pu bugreport, or you can do that if you want.
Note that we ignore test failures on ppc64el which might have caught this
issue.
I don't think so. Tests are not ignored for Bookworm and haven't caught
the issue. OTOH they are ignored for Trixie/Sid, while this version
works fine.
The version command also doesn't trigger the segfault, the autopkgtest
also didn't catch this.
Upstream doesn't care about those architectures, so we're on our own
to resolve issues on architectures other than amd64/i386/arm64. Pretty much
all packages I maintain don't have actual users on non-amd64 architectures,
so I don't consider it worth the effort to ask the porters for help, they
should spend their time on packages that are actually used. With DSA's use
of icinga2 on porterboxes it's the exception to the norm.
Yes, I agree that the upstream situation is not nice. I personally try
to get things fixed [1], but it went nowhere. And the issue was not
really architecture specific, just that icinga2 testsuite doesn't
support page sizes bigger than 4K...
Thanks for your work on that.
Kind Regards,
Bas
--
GPG Key ID: 4096R/6750F10AE88D4AF1
Fingerprint: 8182 DE41 7056 408D 6146 50D1 6750 F10A E88D 4AF1
