Re: -cvs lists oddities

[Jason Molenda <jason-gcclist at molenda dot com>]

Curse you and your amazing attention to detail, Mr. Myers! :-)

Joseph Myers wrote:

> Some questions / oddities about gcc-cvs and related lists:
> 
> gcc-cvs messages say
> 
> Changes by:     whoever@sourceware.cygnus.com
> 
> Surely this should by gcc.gnu.org?

Realistically, either addr is going to work.  It could be fixed
with a good five minutes of conditionalization to the log_accum
script (probably best to check for $ENV{'QMAILHOST'} being set and
using that if it is).  I'm not going to do it.

> gcc-cvs seems to get messages relating to something called "winsup" in
> "/cvs/uberbaum".  Is this a misconfiguration somewhere?

I have no idea what this might be.  'winsup' is a part of the
sourceware:/cvs/src repository, it has nothing to do with /cvs/gcc
repository.

> Why don't wwwdocs changes go through the same log_accum as other changes
> to generate the list messages?  This would add the URLs of diffs to the
> messages.

The scripts for wwwdocs check out the updated pages into a staging
directory and run some HTML preprocesor over them (metahtml IIRC).
log_accum doesn't do any of that.  CVS' CVSROOT/loginfo, which is
where these programs are invoked, is really limited in its abilities,
and one of those limits is that it can only call one script.

No one ever bothered trying to call log_accum from within the web page
updating scripts.  Incidentally, this is related to...

> Changes to libstdc++-v3/docs seem (from looking in CVSROOT) to go only to
> a mysterious undocumented and unarchived libstdc++-webpages-cvs list.  As
> these docs are included in the main GCC sources, shouldn't the messages go
> to gcc-cvs as well?

If you look at the projects hosted on sourceware.cygnus.com, I
don't think any of them publicize their -webpages mailing list or
archive that list.  It's of little interest to anyone except the
maintainers (and those people usually just subscribe to the list).
libstdc++ lives on the sourceware.cygnus.com side of the site, so
it is like all the others.  The setup of the GCC project is odd in
many ways :-), and one of those is that its -webpages mailing list
is publicized and web archived.

In short, the libstdc++ webpages list is not archived because no
one has ever cared whether it was or not.

As an aside, there _are_ archives of the list.  There are _three_
possible archives of every mailing list on sourceware/gcc - the
web archive, the ftp archive, and the ezmlm archive.  ezmlm, by
default, stores every message that goes across a mailing list.  It
uses these archives for a bunch of things, e.g. digest creation
and resending mail messages to a user whose mail address was
temporarily bouncing, so that user doesn't miss any messages.  The
really old messages could be purged from the ezmlm dirs, but I
never did that (I set it up with a lot of free space so this wouldn't
be necessary).  If the web archive and ftp archives disappeared, both
could be recreated from the ezmlm archives if necessary.


Jason