Re: Retiring gccbug script?

["Joseph S. Myers" <jsm28 at cam dot ac dot uk>]

On Fri, 4 Jan 2002, Craig Rodrigues wrote:

> - it does not submit attachments, and this is awkward
>   and time consuming when it comes to analyzing the
>   report in GNATS

It includes plain text testcases for the bug, that can be read directly 
when receiving them on gcc-bugs or gcc-prs.  I might fix a bug when I 
receive it on gcc-bugs if the testcase is plain text; if it's the 
base64-encoded junk (which is not a proper MIME attachment) that gnatsweb 
generates in PRs, I'm not going to investigate the problem or see whether 
it's of interest until much later.

If you want attachments in gccbug-generated PRs, if we move to Bugzilla
then add an attachment facility to the new gccbug script, but make sure
that the resulting gcc-bugs/gcc-prs messages, whatever bug submission
mechanism is used, include the attachments as proper MIME parts,
text/whatever if small or compressed otherwise.

> - it does not give a clear idea how to add more information
>   to a bug report after it has been filed

Reply to the acknowledgment message including gcc-gnats.  The Reply-To on 
the messages is set correctly.  Adjust the text of the messages (and 
gnats.html) if this isn't clear enough.

> - I'm seeing more gccbug submitted bug reports in GNATS that 
>   are corrupted (pending category)

These are not gccbug-submitted - it doesn't let you submit with the
pending category.  These are where people have sent other mail - not PR
submissions or follow-ups - to gcc-gnats.  A human moderator for badly
formatted gcc-gnats mail would be better than filing it as PRs in
"pending".

> What opinions do people have on removing gccbug?
> The guidelines at http://gcc.gnu.org/bugs.html
> do not even mention it.

On the contrary, it is the recommended means of bug submission.  It does 
not create PRs with version listed as "unknown-1.0".  bugs.html clearly 
tells you that the GNATS system is preferred to mailing gcc-bugs, with a 
link to gnats.html; gnats.html clearly tells you that the gccbug script is 
the preferred interface for bug submission.

I don't consider web forms a good means of bug submission.  With gccbug
one can use a text editor for editing bug report text rather than being
constrained into using a web browser, and can submit bug reports offline.
Perhaps the gnatsweb submission form should make clearer that people
should use gccbug if possible.

-- 
Joseph S. Myers
jsm28@cam.ac.uk