One version number, &c, take two

Benjamin Kosnik bkoz@redhat.com
Wed Mar 9 17:16:00 GMT 2005


Hey Zack. I agree that this is a good idea.

> DATESTAMP contains the tree
> datestamp in the usual YYYYMMDD format; it is empty in a release
> tarball.  

Great. This should be all we need, assuming you can continue to update
it in the release tarball (actually, updated on the branch.)

>  - From the libstdc++ folks: Advice on how to eliminate the copy of
>    the date stamp in c++config (note that config/DATESTAMP presently
>    is empty in a release tarball, although that could be changed ...
>    come to think, that would reduce the number of steps in the release
>    process, too).

Yes, this should be changed if at all possible. I don't think this is
unreasonable, actually.

libstdc++ uses this in c++config.h, which is a generated file. Adding
one more piece to the build machinery to paste in DATESTAMP for the
__GLIBCXX__ define doesn't seem especially onerous to me.

I can't do this right now but if you check this in I'll get to it in
the next two weeks.

best,
benjamin



More information about the Libstdc++ mailing list