[Toybox] [PATCH] Enable ASan for github CI.

[enh]

On Mon, Jul 10, 2023 at 11:20 PM Rob Landley <rob at landley.net> wrote:

> On 7/10/23 18:58, enh via Toybox wrote:
> > Untested (because I don't know how to test such changes without merging
> > them), but seems plausible based on the documentation at
> >
> https://docs.github.com/en/actions/learn-github-actions/variables#about-environment-variables
> > ---
> >  .github/workflows/toybox.yml | 2 ++
> >  1 file changed, 2 insertions(+)
>
> Applied, but who is this for? I'm not going to notice: I push from the
> command
> line after testing locally. I check the github web UI less often than I
> still
> check the busybox mailing list.
>

yeah, probably mostly just me. (though obviously others can nag you :-)
there's some magic you can add to the README.md that shows a badge that
says whether or not the tests are passing at ToT that's pretty cool and
surfaces this nicely, but i've no idea how that works.)

but, yes, i actually do usually look at the commits before i kick off a
sync, if only to see "what's changed since last?" and "what commands do i
need to be paying attention to over the next week?". so even if this _only_
helps me, it will help me :-)


> Between your first and second emails I was working to add ASAN=1 to the
> "make
> tests" build target, starting by cleaning up my dirty scripts/test.sh file
> so I
> could check it in (mostly comments, but I've been digging to try to get it
> running under mkroot). But adding ": ${ASAN=1}" to the start of that file
> so I
> could still test without it ala "ASAN= make tests" doesn't scripts/test.sh
> tests
> the existing toybox binary, and for "make tests" toybox is built by a make
> dependency which apparently CAN'T BE TOLD which target we actually called
> on the
> command line, because gnu make is horrific. By the time scripts/test.sh
> gets run
> the value of ASAN doesn't matter, and scripts/make.sh would have to do
> something
> disgusting like ps $PPID to try to figure out that it got called from "make
> tests". Hmmm...
>

ah, that's an even better idea. in that case, maybe revert my change?

especially since we suspiciously haven't had any CI run at all since that
change went in: https://github.com/landley/toybox/commits/master (i'd
expected as least an error, but not even a "ran and failed"!)


> Rob
> _______________________________________________
> Toybox mailing list
> Toybox at lists.landley.net
> http://lists.landley.net/listinfo.cgi/toybox-landley.net
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.landley.net/pipermail/toybox-landley.net/attachments/20230711/96f07a5e/attachment.htm>