Summary: | Segmentation fault with mount_smbfs | ||
---|---|---|---|
Product: | Ports & Packages | Reporter: | dohzono |
Component: | Individual Port(s) | Assignee: | freebsd-bugs (Nobody) <bugs> |
Status: | Closed Overcome By Events | ||
Severity: | Affects Some People | CC: | ardovm, vas, w.schwarzenfeld |
Priority: | --- | ||
Version: | Latest | ||
Hardware: | i386 | ||
OS: | Any |
Description
dohzono
2015-12-11 23:08:20 UTC
Do these symbols exist? /usr/src/contrib/smbfs/lib/smb/nls.c iconv_lib = dlopen("libiconv.so", RTLD_LAZY | RTLD_GLOBAL); if (iconv_lib == NULL) { warn("Unable to load iconv library: %s\n", dlerror()); iconv_loaded++; return ENOENT; } my_iconv_open = dlsym(iconv_lib, "iconv_open"); my_iconv = dlsym(iconv_lib, "iconv"); my_iconv_close = dlsym(iconv_lib, "iconv_close"); Solved. I found that nls.c referes to /usr/local/lib/libiconv.so, so I deintall/reinstalled /usr/ports/converters/libiconv, and this problem has gone. Reproduced. mount_smbfs -E UTF-8:cp936 -N //dohzono@orange/share /mnt/orange with libiconv-1.14_8 => works ok. with libiconv-1.14_9 => core dumped. I have reproduced on 9.3-RELEASE-p33 and libiconv-1.14_9. # mount_smbfs -E koi8-r:cp866 -W sibptus //scanserver@FS03-SIBPTUS/PUB /mnt2 Segmentation fault (core dumped) Need a fix ASAP. I had to force-install libiconv-1.14_8.txz (which I had obtained from a tape dump) and this worked around the problem. Why is libkiconv from the base system affected in any way by /usr/local/lib/libiconv from ports? Shouldn't we reclassify this bug into a more serious category? This may be a fix for 10/stable of this issue. https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=183153 Please see also https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=206740 for a fix. IMHO vas is right; this bug is a particular case of bug #206740. It was triggered by ports r398996, that renamed all 'iconvxxx' symbols back into 'libiconvxxx'. 9.3 is EOL. So i think this is overcome by events. (In reply to w.schwarzenfeld from comment #10) > 9.3 is EOL. So i think this is overcome by events. I think so. And also bug #206740 should be marked as such. |