Stefan Teleman wrote: >> What is the point of the multi-version names? >> I think ld will bind to multi-version name. I don't think you want >> to do this >> on Solaris. > > i think ld will bind to -h'$(SONAME)' which is set in the Makefiles > created by ./configure. in pcre's particular case, (and generally for > open source software which is into the bad habit of using libtool), > the SONAME is usually set to <library-name>.so.0. > > this can be changed in ./configure but: > > - it is a lot of work to change ./configure and configure.in. > - it creates again a fork with patches which won't be accepted upstream > - it will break 3rd party applications which still use libtool (and > the *.la files), and which expect to find this collection of symlinks, > and naming convention for shared libs. > > On the bright side of things, we got rid of the *.la files. OK,... I'm mush less worried (SONAME), but the question still stands: "What is the point of the multi-version name?" Is this just "additional information to the developer"? If so, shouldn't it be Volatile so we can change it at any time? >>> /usr/share/man/man3/pcrestack.3 Uncommitted Manual Page >>> >> Two things here... >> >> By Solaris conventions, shouldn't these all be in man3pcre (or >> something like that)? > > They could be there (this is a verbatim description of the man > documentation as provided by PCRE). This would mean that LibGD2 would > need to have its own man section, CURL will also have its own man > section, etc. Yep. Looked in /usr/man lately? This is what our documentation people wanted and PSARC approved (I forget the case number). I think I abstained from that case, because I disagreed, but it didn't seem important enough to make an issue out of. I really think you need to adhere to this convention.
>> Its a bit strange to see the actual man page in an interface table. >> Can I read this as >> that the interface defined on this page is what is being declared >> Uncommitted? > > The interfaces being declared are indeed Uncommitted, and so are the > man pages associated with them (unfortunately). If PCRE decides (in > some future release we might be interested in) to remove pcreposix and > its corresponding man page, and replace it with pcrerealposix (for > example) and its corresponding man page, then this change will have to > be reflected in the man pages (if we consider that man pages are a > form of interface). OK. I could pink nits about the man page itself needing a commitment. The important thing is that the interface itself has a commitment. I'm good. Thanks for the clarification. - jek3
