On Friday, July 13, 2012 6:24:05 AM UTC+2, Dima Pasechnik wrote:
> The ECL patch > > http://ecls.git.sourceforge.net/git/gitweb.cgi?p=ecls/ecl;a=commit;h=2d639b485023364f215a4acff9300afab8827789 > > you mention in this regard has nothing to do with Cygwin (or at least it > shouldn't), it's there for > the Windows native (or MINGW) support. > I think it does. The problem here is that if WINSOCKAPI is defined, fd_set does not get defined in sys/types.h (have a look in the headers files), so that compilation fails. So it was potentially defined earlier in another place, but on my system it gets defined nowhere. The three 10, 11 and 12 versions fails on my system (outside of Sage, but within it as well), and indeed the git/cvs versions fails for another reason as you mentioned. It's quite strange that you could not reproduce the problem. And I dont think the 32/64 bits stuff is relevant. >From what I saw it's really an header problem. IIRC from what I read on different Cygwin related stuff, time and select and co were (quite) recently modified to ensure that everything is always included, meaning that there is a "circular" inclusion of these headers. The way ECL is buit and these kind of modifications may be the source of the problem. Anyway, removing the inclusion of select.h from sys_time.h solved the problem, i.e. evrything compiled fine (although I did not check that ECL was really functional, except for the fact that it could compute 1+1). That's an ugly workaround, but it will be sufficient to build sage on my system. -- -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org
