On 5 Apr 2003, Alexandre Oliva <[EMAIL PROTECTED]> wrote: > On Apr 5, 2003, Martin Pool <[EMAIL PROTECTED]> wrote: > > > Do we need to build stage1 remotely, or would it be enough to run the > > preprocessor and compiler stages separately for local compilations? > > It doesn't matter how stage1 is built, it's not compared with > anything. It's stage2 and stage3 that must be compiled under very > similar conditions, because they're compared, so that's where we must > not use localhost (because GCC won't generate the same debugging info > when compile a preprocessed file as it would compiling the original > file with the integrated preprocessor, and the preprocessor > maintainers don't want to agree this is a bug :-(
To work with the gcc makefiles does it need to be a command line option rather than an environment variable? Is it much slower to just run everything through 127.0.0.1, rather than introducing a special behaviour? I guess I should measure this. > > Perhaps a DISTCC_SPLIT_LOCAL option would help? > > I can't imagine what this option would do from its name. Is it the > same as the command-line flag I've been asking for? If so, I suppose > it would do just as well. What I meant was that it would invoke the preprocessor and compiler separately for builds on localhost, so the commands executed ought to be the same as is used remotely. > I tried :-( > > The preprocessor maintainers seem to be too focused on performance of > the integrated preprocessor to want to promote programs such as ccache > and distcc, that are built on the assumption that separate > preprocessing shouldn't alter the result of the compilation :-( This is really unfortunate. :-( It's not just that distcc and ccache are useful programs, but it also seems like the right thing to do. -- Martin __ distcc mailing list http://distcc.samba.org/ To unsubscribe or change options: http://lists.samba.org/cgi-bin/mailman/listinfo/distcc
