From: Beman Dawes (bdawes_at_[hidden])
Date: 2001-10-31 10:16:27
At 05:23 AM 10/31/2001, Bryan Scattergood wrote:
>Beman Dawes wrote:
>> At 02:14 PM 10/30/2001, mda_at_[hidden] wrote:
>> >it would be nice if:
>> >1. a release were identical to a snapshot of the CVS at some time,
>> >sans the CVS cruft, without someone's personal build artifacts.
>> Yes, that would be useful. If you would like to submit a script (Bash,
>> Python, or Win2K batch file) that does that, I'll add it to the release
>Why does this need a script? I thought the whole point of "cvs export"
>was to generate a cruft-free snapshot of the repository?
That's an interesting point. The release isn't currently being generated
by export because of fear that that approach wouldn't be compatible with
FrontPage, which is used to manage various aspects of the web site.
But that decision dates from a long time ago, when a lot of things were
I'll take a look at an "export" based approach before the next release.
>> >2. the release history were documented, with a date, scm label,
>> >human-generated release notes, and (ideally) automatically generated
>> >change summary
>> Can't you get that from CVS? If you have a format that is better than
>> CVS output, consider submitting a script to generate it.
>Again, that sounds like an existing tool:
Interesting. But how useful? Does anyone really care that a "Minor fix"
has been made to some configuration file deep in the directory
structure? I guess if it were restricted to some subset (public headers?)
of all the files in the web site, it might be more focused. If you think
one of the available formats would be useful as is, consider running it on
the Boost CVS and posting the output so we can get a look at real output.
There is also a tool chain issue; the release procedure already uses far
too many scripting languages and other tools; adding Perl would make a
minor mess worse.
Thanks for the comments,
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk