On Tue, Jul 28, 2009 at 05:29:23AM +0100, Ganesh Sittampalam wrote: > On Tue, 28 Jul 2009, Trent W. Buck wrote: > > >The patch is already in HEAD, at least. As Joachim says on the > >Debian list, if there's a legitimate bug in zlib 0.5.0 that Darcs > >should avoid, it's better to break haskell-platform compatibility > >than allow compilation with a buggy zlib. > > Just to be clear, to avoid any slandering of zlib ;-) > > zlib 0.5.0.0 is not buggy. Darcs was previously buggy in that it > sometimes wrote out .gz files with bad checksums and then happily > read them back in without checking, so the bug wasn't spotted for a > long time, leaving a significant number of files in repos that had > this corrupt checksum (but were otherwise fine). zlib perfectly > legitimiately refused to read them in, making it unsuitable for use > as the default zlib binding for darcs. zlib 0.5.2.0 adds an > interface to allow clients to be informed of the broken checksums > but still continue, which darcs makes use of to warn the user to run > the new 'darcs gzcrcs' commadn to repair them.
Ah, sorry for the inadvertent slander. With your permission, I will copy your remarks above to the aforementioned Debian bug (#538098). I note that that Darcs currently asks for 0.5.1, but you talk about 0.5.2 above. Is there a typo somewhere? _______________________________________________ darcs-users mailing list [email protected] http://lists.osuosl.org/mailman/listinfo/darcs-users
