[Scummvm-devel] Simon the Sorcerer appears to leak all it's audio ram

Max Horn max at quendi.de
Mon Jun 4 16:21:44 CEST 2007


On Mo, Juni 4, 2007 15:46, Eugene Sandulenko wrote:
> On Mon, 4 Jun 2007 15:33:32 +0200 (CEST)
> "Max Horn" <max at quendi.de> wrote:
>
>> 1) Neil's way, which silently ignores if a failure occurs to load a
>> sound (this change would be inside an #ifdef DS, of course). Kirben
>> is opposed to this, quoting additional bug reports which would
>> supposedly turn up. Kirben, can you please elaborate a bit on this --
>> what kind of bugs would crop up due to this?
> Neil's approach is to do not attempt playing speech when malloc() of
> speech buffer fails. This is really bad and is unacceptable because in
> simon1 CD there are no subtitles. That means that with this hack some
> sentences will be just silently skipped without any knowledge of the
> user, so the user may lost track of what's going on.
>
> Imagine situation when some item description will not fit in DS memory.
> I mean voice over. There are no subtitles. So any user's attempt to
> hear that description will produce no result. Is that good?
>

Of course not, that's why we call it a hack :-). But: Is a crash better,
though? I am still waiting for a viable alternative suggestion.

As it is, though, I can *imagine* situations where a player is confused
because of missing some text. But do those actually occur? So far I am
only aware of this problem occuring in the intro, where missing the text
is completly harmless. Does it occur in any other spot?
And did we get any bug reports over this for the 0.9.x series?


So, yes, this hack is rather evil. But given the short time we have, and
the lack of good alternatives (the only one I heard about came from me,
I'll be happy to hear about more), it seems to still be the best thing to
do. And based on actual experiences made with it so far, the consequences
seem not to be quite as bad as they are painted here.


All the best,
Max




More information about the Scummvm-devel mailing list