On 11/24/2015 08:27 PM, Jeff Epler wrote: > Note: I haven't built these branches, I just looked at the commits. > I think the most important comment is the one on the very last commit.
Thanks for the review Jeff. I accept most of your comments, the ones where i have comments are annotated below. > 174d8f0 handle building under RTAI 4.1 > please review rtai as built to make sure no dumb compiler flags are > specified (optimization flags in particular) I'll double check. > 43083d0 hm2: fix uninitialized variable warning in ioport > would be better to split rename from behavioral change, but don't care > sufficiently much to demand it. My analysis makes me think the warning > is spurious but if your analysis is different please consider for backport Agreed, that would have been cleaner. I'm reacting to a complaint by the Jessie compiler: > /home/seb/linuxcnc-dev/src/hal/drivers/mesa-hostmot2/ioport.c:373:3: warning: > ‘m’ may be used uninitialized in this function [-Wmaybe-uninitialized] Without that commit, if i is never equal to p in count_instances, then m will indeed be uninitialized. This one's a bit half baked, i'll rework it... > ed7603e build system: all versions of RTAI need -msse for math > this has been a source of contention over the years. > May want to check with mbuesch (?) about this. However, I think -msse > is right for FP math on all amd64-type platforms. I can look for mailing > list and irc refs if that is helpful. > > might want to amend to say all *supported* versions, if the point is that > rtai <3 is no longer supported. Ugh... > e8a161a teach build system & rtapi about RTAI 5 > hm, do we need a = "4" branch here? Yep, as soon as someone packages one of the RTAI 4 versions ;-) > 7026610 rtapi: error messages are better than errno numbers > ACK, consider for 2.7? Will do. > 441b9f6 hal: don't segfault if rtapi_init() fails > ACK, consider for 2.7? Will do. Nothing like new flakiness to exercise your failure paths... > b1fbd40 realtime script: when loading, wait for udev to finish > ACK. But if this is a new failure mode, why is that the case? If it > is not a new failure mode, consider for 2.7 or even 2.6. It might just be that my jessie VMs are slower and crappier than the Wheezy and Precise VMs? I don't know, but i did the thing where I instrumented the rtapi_init() function in rtai_ulapi.c and had it system out to 'ls -al /dev/rtai_shm' in a loop, and i consistently caught it with the wrong permissions at the start. I'll arrange this for 2.7 too. After I address your next comment. > fbb72cf realtime script: only try to run udevadm if it exists > I don't think the -x test is right (udevadm needs a path). > Squash to b1fbd40 after correcting. Do real systems not have > udevadm? If this problem was observed in the wild, is the problem > that udevadm is in /sbin and not on PATH or something? If so, > I guess we need a configure test for the location of udevadm? You're right, i'm wrong and my test is wrong. I got confused by this failure: http://buildbot.linuxcnc.org/buildbot-admin/builders/1301.rip-precise-rtai-i386/builds/2866/steps/runtests/logs/stdio Looking at it closer, obviously Precise uses udev, udevadm lives in /sbin, and that script is running without /sbin in the PATH. You're right we need a configure test for the location of udevadm. I'll hack something up. -- Sebastian Kuzminsky ------------------------------------------------------------------------------ Go from Idea to Many App Stores Faster with Intel(R) XDK Give your users amazing mobile app experiences with Intel(R) XDK. Use one codebase in this all-in-one HTML5 development environment. Design, debug & build mobile apps & 2D/3D high-impact games for multiple OSs. http://pubads.g.doubleclick.net/gampad/clk?id=254741551&iu=/4140 _______________________________________________ Emc-developers mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/emc-developers
