-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On 10/5/2010 10:14 AM, Natanael Copa wrote: > Hi, >
Hi Natanael, > The git master have been more or less rock stable since NPTL branch > was merged in (at least for x86). > at least one good point :) > Until the config parser commits. > > The problems are: > * memory corruption due to assuming realloc returns same address as > before. Patches was sent to this list. [1] [2] > * endless loop in getservbyname when service is not known. Patch was sent [3]. > * lines in /etc/protocols and /etc/services (and others?) longer than > 80 chars still gives "interesting" restults. The logic for the realloc > when buffer is too small in getserv*() and getproto*() is flawed and > just don't work. (i can give details if someone is interested) > * memory leaks when calling getservent(), getservbyname() and > getservbyport() multiple times. (fairly simple to fix) > * if you have comments with more than MAXALIASES words "interesting" > things might happen > > Now, I think breaking git master temporary is ok. Its an acceptable > price to pay for progress. I think its also ok that developers gets > busy with other stuff and don't have time to give uclibc 100% > attention. That happens to all. > > However, its not good that git master stays broken. > > So, I have a few options I think we should evaluate: > > 1. Revert the config parser commits, create a feature branch and put > them there, where they can be debugged without stress. > > 2. Raise the line buffer size to 256 chars, which it seems to have > been before. (things that still break at that size would probably have > broken before too). Then rip out the code that tries to be smart by > doing realloc when line buffer is too small. > > 3. Keep line buffer to 80 chars and fix the realloc logic, now. > > Option #2 and #3 implies that the patches[1][2][3] gets applied. > I do not know "config_parser" at all, but I looked your patch times ago and was thinking to push them. I was busy with prelink stuff. > I have some ideas for option #3, but not sure yet if it will work. The > code will be somewhat complicated so I think the extra bytes saved > might get lost in the extra code. > > In any case, I think we should do something *soon*, since current git > master have been broken for too long. I can help but i need someone to > review (sign-off?) and someone with git push access to get the changes > in git master. > Honestly I'm not seeing any activities on uClibc in the last month. I would opt for option #2 and #3. If config_parse needs to be reworked, it can be done as well later. I agree with you to have master fixed asap. I can help pushing your patches [1][2] and [3]. > Thanks! > Cheers, Carmelo > -- > Natanael Copa > > [1] http://lists.uclibc.org/pipermail/uclibc/2010-September/044314.html > [2] http://lists.uclibc.org/pipermail/uclibc/2010-September/044317.html > [3] http://lists.uclibc.org/pipermail/uclibc/2010-September/044316.html > _______________________________________________ > uClibc mailing list > uClibc@uclibc.org > http://lists.busybox.net/mailman/listinfo/uclibc > -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.10 (MingW32) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/ iEYEARECAAYFAkyq82wACgkQoRq/3BrK1s9G9wCgjWeFN0/KIJnrLFmRYxehHbHm 4+0An0DVq9n7mMbChJ4sFd2gZBHQNt5G =mHGh -----END PGP SIGNATURE----- _______________________________________________ uClibc mailing list uClibc@uclibc.org http://lists.busybox.net/mailman/listinfo/uclibc