2013/12/13 Jérémie Courrèges-Anglas <[email protected]>: > Vadim Zhukov <[email protected]> writes: > >>> Vadim Zhukov <[email protected]> writes: >>> >>> > This patch addresses two problems: >>> > >>> > 1) Usage of getdirentries() - seriously. >>> > 2) 64-bit-time_t fix. >>> >>> time_t should be printed with %ll? plus a (long long) cast. Now, >>> I don't see any (long long) cast in the whole OpenCVS code, and GNU >>> cvs uses %08lx and (long) to print the current time... >>> >>> Maybe use %08llx? > > Similar diff for GNU cvs. <Insert joke about CVS and 2038 here> > > > Index: history.c > =================================================================== > RCS file: /cvs/src/gnu/usr.bin/cvs/src/history.c,v > retrieving revision 1.3 > diff -u -p -r1.3 history.c > --- history.c 3 Dec 2013 01:32:49 -0000 1.3 > +++ history.c 13 Dec 2013 15:56:24 -0000 > @@ -866,8 +866,8 @@ history_write (type, update_dir, revs, n > revs = ""; > line = xmalloc (strlen (username) + strlen (workdir) + strlen (repos) > + strlen (revs) + strlen (name) + 100); > - sprintf (line, "%c%08lx|%s|%s|%s|%s|%s\n", > - type, (long) time ((time_t *) NULL), > + sprintf (line, "%c%08llx|%s|%s|%s|%s|%s\n", > + type, (long long) time ((time_t *) NULL), > username, workdir, repos, revs, name); > > /* Lessen some race conditions on non-Posix-compliant hosts. */ > > > [...]
Looks fine for me - but I can't say if we care about upstream compatibility for src/gnu/usr.bin/cvs? If no, we could also zap a useless cast of NULL to (time_t*) here, saving the line length (well, I forgot to mention 2038 here. Oops). :) -- WBR, Vadim Zhukov
