On 5/6/2010 3:21 PM, Dave Korn wrote: > On 06/05/2010 19:24, Charles Wilson wrote: > >> But...if I understand correctly, doing this moves the "fix" for the >> problem Dave identified into that internal code -- which means that the >> mingw guys won't benefit from it, unless they re-implement it in the >> "outside" code. > > The mingw guys won't benefit from this fix until they implement fork > semantics, i.e. never...
Oh, right. You even mentioned the "forkee" right there in the first message after resurrecting this thread. OK, I'll wait to see what cgf cooks up before commenting further. ==Aside: After all my talk about keeping pseudo_reloc.c synchronized, I'm considering adapting Dave's implementation for MSYS, because I'm pretty sure that whatever cgf comes up with for "inside cygwin-1.7.6" isn't going to work very well "inside msys which is kinda sorta cygwin-1.3.4" -- although I could be lucky I suppose. Oh well. There are only a few msys applications/dlls which are built with pseudo-relocs turned on anyway -- just guile and autogen AFAICT -- so "rebuild all of them to obtain new fixes" is not very onerous. -- Chuck
