Re: Rebuilding config_h.SH
[email protected] (Andrew Dougherty) Tue, 4 Jan 2005 08:46:55 -0500 (EST)
| Newsgroups | perl.perl5.build |
|---|---|
| Message-ID | <[email protected]> |
On Tue, 4 Jan 2005, H.Merijn Brand wrote: > On Mon 03 Jan 2005 20:52, Andrew Dougherty <[email protected]> wrote: > > On Thu, 23 Dec 2004, H.Merijn Brand wrote: > > > > > In order to get in sync, I rebuilt config_h.SH with metaconfig, and the > > > content was the same, but reshuffled (reordered) > > > > > > This cause HAS_QUAD to be too far down and also LONGLONG to be used before > > > defined > > > > If I recall correctly, the dependencies in the metaconfig units control > > the ordering in Configure, but not the ordering in config_h.SH. The > > ordering in config_h.SH, I believe, is determined by a simple alphabetical > > ordering of the units. That's what the a_dvisory/ directory is for. I > > think there are some more details in the metaconfig/README file. > > Sounds reasonable, except for the fact that I somehow don't see how and where > a_dvisory is used, if used at all It's not explicitly used -- it's just a name that sorts near the beginning. To be more explicit: when the sorting takes place, the directory name is included along with the unit name, so that a_dvisory/multiarch.U sorts before compline/byteorder.U. > l1:/pro/3gl/CPAN/metaconfig/U 117 > grep a_dvisory * > README:a_dvisory/ > README: The a_dvisory/ directory, then, is a place to put units that need Yup, that's the only place it's documented. > I'm contemplating some sort of tsort algorithm for config_h.SH sorting Yes, something like that would be appropriate -- or use the sorting already determined by metaconfig for the shell portions of the units. As it says in README, however, This could be fixed, I suppose, but it wasn't trivial the one time I looked. -- Andy Dougherty [email protected]