Subject: Re: Applying outstanding patches



Tim Waugh wrote:
On Thu, Apr 14, 2005 at 05:05:42PM +0200, Claudio Fontana wrote:
is repeated in the patch three times.
This cries for a helper function
(convert_multibyte_offset ?).

While factoring out segments like this, it is really important to make
sure that they are entirely the same.

Of course. I used to encounter this in my day job often enough that I wrote an editor macro that showed me a diff between the clipboard and the selected region. I would mark the first chunk, copy, mark the second chunk, diff.

I seem to remember that some of
them are tweaked in various small ways.

Eww. That's a recipe for breakage later.

The patch I currently have
(the one we're talking about) has been well tested, and I don't
believe it introduces any regressions.

Well, good. Obviously your patches have been well tested "in the field" which is greatly reassuring. One of the things I am trying to do with each of these patches, actually, is determine what bug(s) it intends to fix, and to try and make sure we have a regression test for it. Admittedly that may not be practical in the case of "performance" (speed) improvements, but at least for the correctness bugs. I have gone through your patches as logged in the Grep Savannah issue tracker, and tried to add pointers to the original bug reports in the Red Hat tracker, because there are usually explanations of the bug and reproduction recipes there. It's difficult to use a source code patch when the scope and details of the problem it addresses are not specified. If you could go through and add descriptions and pointers to reports to the rest of them, that would be great.

Feel free to restructure it, but do please make sure to test it very
thoroughly.

Sure, but maybe you can help us with testing. For instance, did you use any automated test scripts?

I think "more harm than good" is a little melodram...

atic, don't you? :-)

Well, maybe a little, but actually I would concur that avoidance of such duplication is extremely important and that some of the bugs that already exist can almost be blamed on poor code structure - in that it would have been much easier to spot or avoid them if it had been well structured.

- Julian

--
http://www.foad.me.uk/







Privacy