Re: Update on bugzilla stuff

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

On Tue, 1 Jan 2002, Daniel Berlin wrote:

> I removed the platform and os fields, fixed some other bugs/missing things 
> in conversion i noticed/you guys pointed out, and added the 
> host/build/target fields and extraction of them from the prs.
> It also puts the environment field contents into a comment.
> I had this done a week ago, but was out of town, and it's all on a new 
> machine (installed today), so i didn't get a chance to put it up till now.

Suspended PRs are still mapping to RESOLVED/LATER - as Neil noted, this is 
wrong, the status of such PRs should be something that does not indicate 
them to be resolved, such as NEW/LATER, and which is found in the default 
search parameters.

The search form still looks a mess in Lynx, with multiple fields (Product,
Component, Version) having their values in a single vertical list with all
the headings above that list.

The versions listed should distinguish 2.96 from 2.96 (redhat), and 
probably 3.0-pre from 3.0.

Some bug reports (e.g. 545) still get links such as "mailto:neil"; without
an @gcc.gnu.org.  Also, that bug report has Reporter listed as 
gnats@gcc.gnu.org, which is wrong.

> Attachments are compressed in the database (using Compress::Zlib, of 
> course), since most of the time they are huge preprocessed text files.
> This trims the attachment database from 210 meg to 63.
> This also prevents creating MySQL full-text search indexes (unless i 
> hacked mysql to handle searches on compressed text, which i'm not going 
> to), but i figured this was likely not going to be useful for attachments 
> anyway.

When we get attachments on new bugs going out by mail with the bug reports
to gcc-bugs and gcc-prs (as well as new attachments on old bugs also going
out by mail), small ones should probably be sent out as MIME text/plain
attachments to the messages (if indeed plain text), larger ones compressed
in mail.  That attachments aren't normally compressed is one of the
problems with the current GNATS system.

> There is one bug report that was too large to insert into the 
> database without fiddling mysql parameters,  which is 5154. 
> This happens because the file is not a real attachment, so it wants to 
> insert it as a comment. The query times out inserting 21 meg of 
> text.  This is not a server or client bug, it just takes that long to 
> un-sql quote it,parse it, etc.
> Right now I handle this bug report manually (the perl script doing 
> the conversion specifically skips it), since 8 bazillion pages of comment 
> text would be useless anyway.

Making the comment text into an attachment manually in the conversion
script for this PR is probably the appropriate solution.

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