TRACE(...., debugstr_w(uninitialised data))?

Hi,

winmm/mci.c contains some code like this:
TRACE("=> 1/ %x (%s)\n", dwRet, debugstr_w(lpstrRet));

However, in case of error, the lpstrRet buffer is likely not initialised.
This could cause
 1. debugstr_w to choke on bad/unconvertible data (not even 0-terminated);
 2. Valgrind errors (I don't know whether the Valgrind warnings I observe
    are due to this line of code).

Am I worrying too much or should any such piece of code be protected by
TRACE("=> 1/ %x (%s)\n", dwRet, LOWORD(dwRet) ? "" : debugstr_w(lpstrRet));
(actually, an even more convoluted test is needed because of MCI_INTEGER_RETURNED)

Or is it not worth careing about because it's within TRACE() only?

I assume that debugstr_w is not executed when that channel is not traced. Correct?

Regards,
 Jörg Höhle

[Juan Lang]

Hi Joerg,
> Am I worrying too much or should any such piece of code be protected by
> TRACE("=> 1/ %x (%s)\n", dwRet, LOWORD(dwRet) ? "" : debugstr_w(lpstrRet));
> (actually, an even more convoluted test is needed because of MCI_INTEGER_RETURNED)

I'd worry about it.  We've sometimes had bugs where a crash appears in
a failure mode when a particular debug channel is enabled, but that
debug channel is the very thing that might help us debug the failure
better.
> I assume that debugstr_w is not executed when that channel is not traced. Correct?

That's correct.
--Juan