On Sun, Jul 27, 2014 at 10:46:53AM -0500, Bruce Dubbs wrote: > Ken Moffat wrote: > >On Sun, Jul 27, 2014 at 11:45:46PM +1200, Christopher Gregory wrote: > >>Hello, > >> > >>I have just had confirmation that a patch that I located to fix run-time > >>segfaults when mounting remote shares using nfs-utils-1.3.0 compiled > >>with gcc 4.9.x, works. > >> > >>I have left the detailed instructions describing what the original bug > >>submitter had in it, as it describes the error in detail and the > >>responses from three red-hat developers. > >> OK, I've now read the patch, and it does seem to fix a potential problem. If nobody beats me to it, I'll take a look this week to see what upstream are doing, or failing that, the distros. Christopher put me on the back foot by saying the patch came from a gentoo message board - their useful patches usually get into their ebuilds.
But I would still feel more comfortable if somebody can explain in what circumstances the original problem happens. > >>If someone would like to add it to the nfs-utils page in TRUNK, I think > >>that it could help a number of others. > >> > >>Regards, > >> > >>Christopher. > >> > > > > At the moment, I disagree - it seems to be a workaround for > >building with -Os. I don't have a problem with the patch being in > >the patches repo - we have various other patches which aren't in the > >book. And looking back, I must have been getting mixed up with something else - I see that our report is for -O2, but it ran ok with -O1. I'm not sure which version of gcc was involved (4.9.0, or 4.9.1?). And which version of nfs (2 or 3 ?). So, I still do not understand why the unpatched version works for me (unless it is a new problem with gcc-4.9.1) but I'll take a look at this in the next few days. > > > > For me, the client part of nfs-utils-1.3.0 is working fine (nfs v3) > >with gcc-4.9.0. I build with -O2 whenever a package respects my > >CFLAGS, in this case the default flags are -g -O2. > > > > Using -Os used to be a nice idea (back in the days when a desktop > >machine might only have 512MB or RAM). For a while, Linus thought > >that using -Os in the kernel would be a good idea (to reduce the > >icache footprint), but too many problems showed up. So, I recommend > >that people not use -Os in normal circumstances. > > I generally recommend using whatever upstream uses as the default. That's > what they have tested the most. Size should not be an issue today with the > tiny price of RAM/Disk. Speed differences are imperceptible at the speed of > modern systems. > > That said, stripping out debugging symbols is a reasonable thing to do, but > that can be done any time. > > -- Bruce > -- > http://lists.linuxfromscratch.org/listinfo/blfs-dev > FAQ: http://www.linuxfromscratch.org/blfs/faq.html > Unsubscribe: See the above information page -- Nanny Ogg usually went to bed early. After all, she was an old lady. Sometimes she went to bed as early as 6 a.m. -- http://lists.linuxfromscratch.org/listinfo/blfs-dev FAQ: http://www.linuxfromscratch.org/blfs/faq.html Unsubscribe: See the above information page
