Hi Dan,
(re-cc'ing wine-devel)
Running it through the buildbot now... seems ok on first two tries, so you may have accidentally fixed the problem.
I'll resubmit to wine-patches shortly, probably tonight.
I ran the stuff through the bot, 0 errors everywhere except on NT (NT fails totally, do I really need to investigate that?).
Yes, you may need to skip these tests on nt.
I don't have a i7 machine,
It's probably just mine that gets lucky. Some tests fail better on some machines than others. Did you test with WINEDEBUG=warn+heap ?
No, I didn't. I thought you wree telling me that that made your problem go away, so I didn't bother. I can do that of course if it helps.
but I tested on a Westmere EX server I have access to, no problems with the conformance test. I also did some valgrind runs on that machine, but I have little experience in reading valgrind output.
Did you follow http://wiki.winehq.org/Wine_and_Valgrind ? Feel free to ask questions.
Yes, I did. I am seeing no regressions (read: additional warnings and/or errors) between the trunk and my patched version.
A few notes:
if (!in_quotes && (*p == ' ' || *p == '\t')) {Could you use isspaceW() ? I suppose not in the case statement below, but you check this in several places, and the function might be clearer (if it's correct)?
Honestly, I didn't consider it. I tried to change as little as possible of the existing code.
Is there duplication between WCMD_parameter and WCMD_parse that could be gotten rid of productively?
Yes, there is, as I noted in an early posting on this subject. Indeed the two functions have essentially the same purpose. However I don't feel bold enough to try fixing that as long as my current stuff isn't accepted.
Originally, all that I wanted to fix was 'echo "Hi"'. I have gone quite a few steps ahead from there with my current patch set. When I see this getting merged I'll consider further fixes, but I won't be able to follow up *much* further.
Best regards Martin