On 2008-09-03 01:39, René Berber wrote: > Brandon Perry wrote: > > >> IIRC, the announcement said that POSIX breakage would happen if you tried to >> compile 0.94 on cygwin... >> > > If you mean this part : > > ** This version drops the special support for Cygwin. Our QA process showed > ** serious problems with ClamAV builds under Cygwin due to some low-level > ** incompatibilities in the POSIX compatibility layer, resulting in > unreliable > ** ClamAV behaviour. > > It sounds like non-sense to me, version 0.94rc1 works fine, and both > versions build fine. The above statement seems to come from somebody > that didn't even test, didn't look at what 'make check' was doing, and > didn't really care. > > "serious problems" ? Only problem is the test I mentioned, passing a > file descriptor is not supported under Cygwin as far as I know. >
No, file descriptor passing is not a 'critical' feature, and is not the reason for dropping Cygwin. Passing file descriptors among arbitrary processes can't work on Windows because it is not supported by the OS, it only works on systems supporting Unix domain sockets. [of course file desc passing works across fork(), but that is not the point here] This is not cygwin's fault, and there isn't anything cygwin could do about it. > "unreliable behaviour" ? Does that mean the distribution/use of older > bzip2 libraries? The warning at configure time is clear, I'm not sure > it is correct, the CVE says "before 1.0.5", Cygwin uses 1.0.5 . > > And the gem: "incompatibilties in ... compatibility layer"; it is or it > isn't compatible? The Cygwin folks make an effort for being POSIX > compatible. > Have a look at bb #1139 for an example: lstat() failing for no apparent reason. Try 'clamdscan ../test/clam.exe' vs. 'clamdscan clam.exe'. Best regards, --Edwin _______________________________________________ Help us build a comprehensive ClamAV guide: visit http://wiki.clamav.net http://www.clamav.net/support/ml
