[Pkg-libvirt-maintainers] Bug#980334: libvirt FTCBFS: uses the build architecture compiler via dtrace

[Andrea Bolognani]

On Sun, Jan 17, 2021 at 10:58:39PM +0100, Helmut Grohne wrote:
> Source: libvirt
> Version: 6.9.0-1
> User: debian-cross at lists.debian.org
> Usertags: ftcbfs
> 
> We're getting closer to making libvirt cross buildable. A lot of
> libvirt's dependencies have been fixed. One aspect that resides in
> libvirt is dtrace. dtrace runs a compiler and defaults to the build
> architecture compiler. As such it'll fail finding host architecture
> headers and that happens to break the build. When running dtrace, one
> must export the CC variable with a suitable value. Implementing this
> with meson is not entirely trivial due to
> https://github.com/mesonbuild/meson/issues/266. The attached patch
> therefore implements the suggested workaround. Please consider applying
> it to bring libvirt one step closer to being cross buildable.

Hi Helmut,

I was initially surprised by your report of FTCBFS, since the libvirt
project performs cross-builds on Debian as part of the regular
upstream CI pipelines.

After looking more closely, however, I realized that we're not
installing the foreign systemtap-sdt-dev package in the environments
used for cross-builds, whereas it's listed among Build-Depends in
Debian. So that explains why the upstream pipelines never caught
this issue.

This is a pretty obvious gap in upstream CI coverage, and we should
take the opportunity to address it. When I do, the result is

  https://gitlab.com/abologna/libvirt/-/pipelines/243473913
  https://gitlab.com/abologna/libvirt/-/commit/f6184ca2629776b935e37f4f6e1acee4564b7843

just as one would expect based on your report; applying your patch on
top results in

  https://gitlab.com/abologna/libvirt/-/pipelines/243475989
  https://gitlab.com/abologna/libvirt/-/commit/dd0cc0ab8c42e78be56f2939415b096f916e3e77

which is much better :)

Your patch looks reasonable and appears to work, but I'd rather have
it go through upstream, with the additional eyeballs that entails,
and then cherry-pick it into Debian instead of the other way around.

Are you okay with me submitting the patch upstream on your behalf? I
will of course make sure that authorship information are preserved
and that you get sole credit for your contribution.

The patch, as I intend to submit it upstream, is attached. Please let
me know if I can go ahead, and thank you for not only reporting this
issue but going the extra mile and providing a patch that fixes it :)

-- 
Andrea Bolognani <eof at kiyuko.org>
Resistance is futile, you will be garbage collected.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0001-meson-Fix-cross-building-of-dtrace-probes.patch
Type: text/x-diff
Size: 2564 bytes
Desc: not available
URL: <http://alioth-lists.debian.net/pipermail/pkg-libvirt-maintainers/attachments/20210119/da27440c/attachment.patch>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: not available
URL: <http://alioth-lists.debian.net/pipermail/pkg-libvirt-maintainers/attachments/20210119/da27440c/attachment.sig>