[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Patch for

From: Lars Henriksen
Subject: Re: Patch for
Date: Sun, 29 Sep 2002 21:02:43 +0200
User-agent: Mutt/1.4i

On Wed, Sep 25, 2002 at 08:14:14PM +0200, Lars Henriksen wrote:
> On Wed, Sep 25, 2002 at 05:59:53PM +0200, Yngve Svendsen wrote:
> > At 17:39 25.09.2002 +0200, Lars Henriksen wrote:
> > >Do we want to (must we) keep backward combatibility?
> > While working with the recent "Sort By" patches, I had occasion to look 
> > more closely at this. Compatibility with stored Gnatsweb 2 queries is a 
> > non-issue I think. There are a few sites I know of that use Gnatsweb 3.99.x 
> > in production (Dirk, this probably concerns you), but if we are ever to rid 
> > Gnatsweb of this totally unnecessary level of complexity, this is the time.
> > I think I'll use my Supreme Maintainer Authority to decree that field2param 
> > and param2field must die.
> Fine. That implies that the query expression logic in submitquery() must be
> corrected since it depends on case differences.
> > Relevant patches are very welcome, since I have somewhat limited time to 
> > code right now

Attached is my attempt at eliminating the field/parameter/upper-case/lower-case
conversions from gnatsweb.

Our friends field2param() and param2field() have gone: gnatsweb uses dbconfig
field names exclusively (this is somewhat ironic since gnatsd happily accepts
category for Category, etc.).

With this patch gnatsweb is NOT backwards compatible, neither with versions 2.* 
nor 3.*:

1) Stored and bookmarked queries may no longer work or may return different
   results (all field names are now exact dbconfig field names).
2) submitquery() uses anchored regexp searches (^ and $) for enumerated
   types. For list box selection this leads to the expected result; for
   regular expression queries (advanced queries) it still leaves the
   possibility of inserting suitable ".*". This isn't the best possible
   solution, but one requiring only a small change in the code. I prefer
   postponing a better one (divide and conquer).
3) Field related callbacks (sendpr() and edit()) must have their "reason"
   adjusted to use dbconfig field names.

The patch is against the current HEAD (revison 1.101).

Lars Henriksen

Attachment: xx
Description: gnatsweb context diff

reply via email to

[Prev in Thread] Current Thread [Next in Thread]