Dan Kegel wrote:
Alexandre Julliard wrote:
This can't work, the sprintf is supposed to modify the buffer, not a copy of it.
Also note that registry functions are not supposed to set last error, so you probably don't want to use Dmitry's wrapper here.
Hmm. Would it be possible for janitors to as a matter of course submit a regression test program for the routines they are sweeping up? That might avoid a few of these kind of issues.
- Dan
I don't agree. The way I see it these janitorial projects are supposed grunt work, and a lot of these W->A changes are just that. They are just simple changes that anyone can do. Making it (writing the test) a requirement might raise the bar to the point where no one (else) will even attempt to do this stuff.
I was not comfortable with the sprintf situation when I wrote it and I suspected it was wrong. The best way, that I know of, to get these problems resolved is to get someone else to look at them, and submitting a patch is the best way to get the feedback I need. The bottom line is that, even if we had conformance tests I would still be up against the sprintf/swprintf problem.
OTOH I was going down the list of W->A calls and tried to work on dlls/advapi32/crypt.c It has no tests and the amount of problems with it hurt my brain. I have no intention of going any further with it without some conformance tests. Even with conformance tests, I have my doubts about being able to do without at least some feedback.
IMO after the simple changes are done, some of these W->A projects (like crypt.c) are going to require some detailed knowledge. Exactly what to do with them I don't know but I would not think of them as simply janitorial.
I think it would be a better idea to have adding missing tests as a separate janitorial project. Perhaps Patrik Stridvall has a inclination to help out here. (So we have a list)