<similar to Bug #86437, but that's on RHL 8.0 where glibc is compiled differently> On an Athlon running RHL 9 + glibc-2.3.2-23.9, I see the following: wine-20020605 from RHL 8.0 dies for me, whether recompiled or using the 8.0 binary, and whether LD_ASSUME_KERNEL is forced or not: [cricker@winsucks cricker]$ winemine wine: chdir to /home/cricker/.wine/wineserver-winsucks : No such file or directory [cricker@winsucks cricker]$ LD_ASSUME_KERNEL=2.2.5 winemine wine: lstat /home/cricker/.wine/wineserver-winsuck/socket : No such file or directory [cricker@winsucks cricker]$ Crossover Office 1.3.1 dies for me, whether LD_ASSUME_KERNEL is used or not wine-20030318 (latest release) gives the following warning when compiled from source: config.status: executing windows commands *** Warning: non-reentrant libc detected. Wine will be built without *** threading support. Consider upgrading libc to a more recent *** reentrant version of libc. but then compiles and runs
Without LD_ASSUME_KERNEL wine is certainly not expected to work (winehq folks need to hack support for NPTL first). Have you tried LD_ASSUME_KERNEL=2.4.1 ? That's what you were using in RHL 8.0, unless LD_ASSUME_KERNEL=2.2.5 was used when running wine. When you're compiling wine yourself, it is expected that it doesn't find a pthread implementation, unless you run wine build with LD_ASSUME_KERNEL=2.4.1.
Looking at wine more closely, libwine.so attempts to override __errno_location, __h_errno_location, etc. This is not going to work in glibc 2.3.2 which does bypass the .plt call to these, as it has separate copies of this function in libpthread and libc. Guess this is something you should take with the wine/crossover folks. They know about NPTL and perhaps are working on support for it, but they should know gilbc 2.3.2+ linuxthreads don't work either.