On 12 March 2018 at 11:25, Daniel Stone <dan...@fooishbar.org> wrote: > Hi, > > On 12 March 2018 at 11:21, Emil Velikov <emil.l.veli...@gmail.com> wrote: >> On 9 March 2018 at 11:09, Daniel Stone <dan...@fooishbar.org> wrote: >>> On 9 March 2018 at 10:59, Emil Velikov <emil.l.veli...@gmail.com> wrote: >>>> - above all, the internal path is a 'dummy' fallback. anyone can >>>> provide the binary name as an argument >>>> $ .../wayland-egl-symbols-check .../libwayland-egl.so >>>> - since we have a fallback - a plain .../wayland-egl-symbols-check >>>> will work most of the time >>> >>> That makes sense, running it from the build root. Is that just because >>> 'make check' is slow, or? (sanity-test is really slow.) >>> >> Short back story: I was playing with OBS and getting the build >> artefacts (the contents of a failing test) was a pain. >> Admittedly, there may be another way to handle that, although in general: >> >> Passing the file as argument makes debugging a lot quicker/easier. > > Sure, that makes sense. I have a suggestion though: why don't we try > argv[1] first, then fall back to $WAYLAND_EGL_LIB, and error out if > neither are set? That way we don't hardcode internal autotools > implementation details into our own scripts (remove fallback > definition), allow autotools tests to run easily (environment > variable), and allow people to use it manually for debugging > (command-line argument). For the latter, I think being explicit is > better: it means you can't accidentally forget to set an environment > variable and end up testing the wrong thing. > In all honesty keeping things simple was my priority. I was hoping for some compromise, but it seems like my allowance is out :-(
Modulo any objections, I'll just respin the series to drop this patch all together. People are open to hack things up when they hit a problem. -Emil _______________________________________________ wayland-devel mailing list wayland-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/wayland-devel