Regression test for thread safety?

Loren James Rittle
Mon Jul 15 21:12:00 GMT 2002

> Something you should know if you haven't discovered it since your email:
> turning thread support on and off is only supported by the command-line flags
> like "-pthread" or "-mthreads" or whatever is appropriate for your platform.
> Preprocessor symbol hacks like _NOTHREADS were never documented and aren't
> really supported.

Phil, what you say about being undocumented is completely true for all
early GCC releases that included SGI STL.  What you say about using
"-pthread" or "-mthreads" or whatever verses defining _NOTHREADS is
finally true in GCC 3.1 and beyond.

GCC 3.0.X still allowed a user to define _NOTHREADS.  Dan was asking
about 3.0.2.  There are known SMP bugs in that release.  Due to my
fault, I don't think they were all ever fixed on the 3.0.X branch.

Loren J. Rittle, Principal Staff Engineer, Motorola Labs (IL02/2240), KeyID: 2048/ADCE34A5, FDC0292446937F2A240BC07D42763672

More information about the Libstdc++ mailing list