On Fri, Aug 21, 2015 at 1:46 PM, Chad Versace <[email protected]> wrote: > On Sat 15 Aug 2015, Ilia Mirkin wrote: >> Signed-off-by: Ilia Mirkin <[email protected]> >> --- >> src/waffle/gbm/wgbm_display.c | 8 ++++++++ >> 1 file changed, 8 insertions(+) > > Hi Ilia, > > The patch, as-is, silently falls back to searching udev if it fails to > open the device path given in the environment. I think that's an > unexpected and surprising failure mode, because the user thinks one > device was opened (the one in the environment) but is given no > indication that a different device was opened. > > Like we discussed in IRC, I made a few changes to the patch, and also > fixed the silent failure mode. It's in master now [1]. And I'll add it > to 1.5.2 too (I'm preparing the release notes right now). > > [1]: > https://github.com/waffle-gl/waffle/commit/cef67496e379b660601f64ab2ddf7cf06307b8aa
Even better! I wasn't a huge fan of the fact that you had to specify things that udev understood vs an actual device node. (But I also didn't want to make too invasive of a change due to lack of familiarity with the code.) It's questionable whether the override should only happen when name is null, I could see the argument either way, but I don't really care since my only use-case for waffle is piglit, which passes NULL. Thanks again for taking care of this. -ilia _______________________________________________ waffle mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/waffle
