Re: Parsing mail to Bugzilla

[Daniel Berlin <dberlin at dberlin dot org>]

On Tuesday, March 25, 2003, at 04:23  PM, Tom Tromey wrote:
> > > > > > "Dan" == Daniel Berlin <dberlin at dberlin dot org> writes:
>
> Dan> Number is unique in bugzilla.
> Dan> It ignores component when parsing.
>
> I think it would make sense to loosen the allowed formats for
> mentioning a bug in a commit message.

My current log_accum (and bugzilla followup handler, though without the 
last little bit since it has no log message to deal with in a subject) 
regex looks like:
m/(?:Bug|PR)(?:\s*)(?:[a-z]+\/)?([\d]+).*/si

IE it can handle Bug|PR <any amount of space> <optional word 
prefix+/><number><rest of log message>, case insensitively.


The only thing it verifies is that the bug id matches the product 
passed to log_accum IE that bug <number> is really for product 
<whatever was passed on the command line>.
This was done so that if other projects want to use bugzilla, their 
log_accum won't pick up non-project bugs accidently in commit messages.

Though this might make sense anyway, thinking more about it (this would 
mean that if gdb switched to bugzilla, and a gdb commit mentioned a gcc 
bug, or a gcc commit mentioned a gdb bug, the bug would get a comment 
appended about the commit).
T
> Doesn't bugzilla parse the text
> of bug entries, looking for references to other bugs?
Yes. Not just references to other bugs, but references to attachments 
in other bugs, comments in other bugs (IE you can directly link to a 
comment in another bug), etc.

This is done in bugzilla's bug display stuff, not log_accum's "is this 
for a bug" time.

>  Perhaps we
> could allow all those styles as well.
>
> Tom

> Tom