Re: [PULL 00/51] Build system, i386 changes for 2023-09-07

[Kevin Wolf]

Am 11.09.2023 um 12:22 hat Philippe Mathieu-Daudé geschrieben:
> On 11/9/23 12:10, Philippe Mathieu-Daudé wrote:
> > On 8/9/23 17:47, Stefan Hajnoczi wrote:
> > > I wonder how it passed CI?
> > > https://gitlab.com/qemu-project/qemu/-/pipelines/996175923/
> > 
> > The conditions are:
> > - x86 host

I didn't actually try other hosts, but why do you think it depends on
the host? Is it not x86 target?

> > - both system / user emulation enabled

I did try this one while bisection because I didn't want to build the
system emulator in each step. User emulation only is enough, you don't
need to build the system emulator to get the build failure.

> > - KVM disabled

I didn't explicitly disable KVM, but I suppose it's automatically
disabled for linux-user.

> > - debug enabled
> 
> Oh, I forgot:
> 
> - clang compiler
> 
> (because this isn't an issue with GCC).
> 
> > We have jobs with 3 of the 4, but none with
> > all the 4.
> > 
> > Is it worth test it?

So it's x86, user emulation is built, debug enabled, on clang. This
doesn't feel very exotic, though of course --enable-debug will be more
common for developers while it's not the primary test case for CI.

I don't see any existing test to which it could be added without losing
another important configuration. Not sure if it's worth an additional
test when we're already hitting the CI limits too often.

On the other hand, now that I'm looking at the CI configuration, is it
correct that we have no build for clang + --enable-debug at all? Even if
you ignore the other conditions? Maybe that's basic enough that we
should have it.

Kevin