RE: Subject line processing in Gnats 4.0

[Dirk Bergstrom]

i've been wrestling with this problem for some time (we've been running
4.0a for over a year...), and i've been meaning to send this to the list
for a while:

summary:
at my company, we have been having some difficulties with gnats' 
handling of email replies to PRs.  gnats checks the subject of incoming 
mail, and if it matches a fairly restrictive pattern (must look like a 
reply), the message is appended to the relevant PR.  if there is no 
match, a new PR is created.  this often leads to creation of bogus PRs, 
because the matching expression is too restrictive.  i propose that we 
remove the restriction, and append *all* messages with subjects matching

a PR to that PR.

detailed explanation:
when a message is filed by queue-pr (via file-pr), gnats scans the 
subject line for the following regex (line 575 in file-pr.c):

"(.*re[ \t]*(\\[[0-9]+\\])?:)?[ \t]*([-a-z0-9_+.]*[:/][ \t]*([0-9]+))"

which (sort of) translates to a regex of /.*(re:)? <category>/<pr_num>/.
if it 
finds this pattern, the message is appended to pr_num; if not, gnats 
creates a new PR, using the subject as synopsis, and body as
description.

this only works if people using gnats pay close attention to the subject

of their emails.  in the real world, however, developers forward 
messages, they use mailers that insert extra cruft in the subject line 
(email addresses, names, etc), and sometimes gnats is used in concert 
with a different bug-tracking system, which inserts it's own tracking 
number in the subject.  whatever the cause, the result is a bogus PR, 
which clutters up the system, confuses developers and support staff, and

takes time and energy to close/expunge.

these subjects do the Right Thing:

*) re: compiler/9876: compiled code runs backwards
*) serial-port/1234: cannot process frosted flakes

but these subjects all do the Wrong Thing (create new PR, instead of
appending):

*) Re: FW: sw-kern-sanders/16543: kernel panic on gunshot
*) PR 12345
*) Fw: sw-rip-vnwnkl/10795: The routing subsystem is asleep
*) RE: 2001-0821-021: hw-pr0n/16413 FPC shows pictures of naked ladies
on reboot

(basically, anything in front of the category that's not "re:" bombs.)

i believe that the current behavior violates the principle of least 
surprise -- if i send a message to bugs with a subject "re: fw: 
sw-foo/1234", i expect that gnats will append it to PR 1234.  opening a 
new PR is *not* what i intend.

solution:
it's not clear to me why the regex is so restrictive.  i think it would 
make more sense for gnats to assume that any message with a parseable PR

identifier (<category>/<pr_num>) in the subject should be appended to 
that PR.  that would allow for all the real-world messages to be 
processed in a manner that would not surprise the sender.

furthermore, i think it should also accept "PR<pr_num>", if it is the 
first reasonable text in the subject, a regex along the lines of "^[ 
\t]*((re|fw):)?[ \t]*pr([0-9]+)".  i often see bogus PRs with this 
synopsis, which were clearly intended to be part of <pr_num>, not 
separate PRs.  however, this would be a more extreme change, and might
have unforseen consequences...

does anyone see a reason to keep the restrictive regex?

--
Dirk Bergstrom               address@hidden
_____________________________________________
Juniper Networks Inc.,          Computer Geek
Tel: 707.433.0564           Fax: 707.433.0769

> -----Original Message-----
> From: Milan Zamazal [mailto:address@hidden
> Sent: Saturday, December 08, 2001 3:23 PM
> To: Michael Richardson
> Cc: address@hidden
> Subject: Re: Subject line processing in Gnats 4.0
> 
> 
> >>>>> "MR" == Michael Richardson <address@hidden> writes:
> 
>     MR>   Following the code through, it calls sends SUBM to the
>     MR> server... and there is some heuristic in the gnatsd that sees
>     MR> the subject line. And this has worked... sometimes. 
> But I have a
>     MR> number of times it hasn't worked, and new PRs were created. I
>     MR> can share them if this isn't a known problem.
> 
> The problem is not known to me.  Please provide me with the details.
> 
> Thanks,
> 
> Milan Zamazal
> 
> -- 
> It's amazing how much better you feel once you've given up hope.
>                                                 (unknown source)
> 
> _______________________________________________
> Help-gnats mailing list
> address@hidden
> http://mail.gnu.org/mailman/listinfo/help-gnats
>