catch all the possible reallocs that could return a NULLto later be…

[psi29a]

…dereferrenced

Try catching issues raised here:
#192 (comment)

[psi29a]

@sezero any thoughts or comments?

[sezero]

• WM_ERR_CORUPT seems wrong. Should possibly be WM_ERR_MEM (or whatever is there for it.)
• exit() from the lib is BAD: some error return from the relevant procedure is needed.
• upon realloc() failure, old ptr is not free()d and leaked. we need maybe something like:

  type *tmp = (type *) realloc(buf, newsize);
  if (!tmp) {
   /* handle error */
  }
  else {
    buf = tmp;
  }

[psi29a]

I'm heading off to bed, feel free to commit into this pull request if you see something that should be fixed. I'll catch up with this tomorrow. Thanks for reviewing.

update: Back...

[sezero]

cb85b69: Made it to compile c89-only compilers.

I still don't like this. We are not really handling errors. Not really.
See, e.g. _WM_CheckEventMemoryPool(): it just returns, no one
checks failure,and for this particular case the old pointer still is
not freed.

This needs more more work, and I believe it isn't that urgent.

[psi29a]

No, it isn't urgent. If a library can't reallocate memory, then we're screwed anyway. A null dereference will cause a seg fault (or something similar in windows), which is worse than having a library just 'exit'. So I can understand 'logging' the issue, but we still need to warn the application about the issue so that it can be aware of the issue.

[psi29a]

This PR is ancient.. not sure if it's still worth it or will need re-work later? Otherwise I close

[sezero]

This PR is ancient.. not sure if it's still worth it or will need re-work later? Otherwise I close

As you wish.

[sezero]

I rebased this myself (and even did a minor correction.) It compiles but
suffers from returned value check issues:

• internal_midi.c: _WM_CheckEventMemoryPool returns void, therefore
  the failure isn't known to its callers. (some of its callers do return
  a value themselves, but do their callers check in return??)

• patches.c: _WM_load_patch returns void therefore the failure isn't
  known to its callers.

• mus2mid.c,xmi2mid.c: resize_dst returns void therefore the failure
  isn't known to its callers.

If you are interested in this patch, those issues need to be resolved.