On 02/01/2011 10:50 PM, Paul Eggert wrote: > On 01/31/11 23:23, Markus Duft wrote: > >> [1] http://bugs.gentoo.org/show_bug.cgi?id=352254 > > Ah, OK, so it's a bug in the underlying file system implementation. > Unfortunately the patch proposed in > <http://bugs.gentoo.org/attachment.cgi?id=261159> is bogus, > as it introduces a race condition and a potential security hole.
Thing is, it worked in 1.23 (not 1.24) flawlessly both on interix and linux, so i don't think it's "only" a bug in the filesystem (although i'm with you: NTFS is a buggy fs ;)). also the strange time-related component while debugging is not explained by this. i suspect that a change in gnulib is the actual problem, but i was completely unable to find a hint. I use NTFS on interix, and i don't know what darkside uses on linux, where the patch fixed the issue too (see the bug, i asked him about the filesystem). Any other idea, apart from "filesystems fault"? I'll happily try to debug things, however for some reason the "bootstrap" script from git does not run when i checkout tag 1.23, but only with tag 1.25 ... i plan to try and bisect gnulib in the 1.25 tar build to get a better idea of whether there is a problem in there. markus > > Can you find out more about the file system bug? Perhaps we > can test for it dynamically, and warn users about it, or something > like that. >
