[Scummvm-tracker] [ScummVM :: Bugs] #14892: DOTT: Crash when trying to play MM on Ed's computer

[ScummVM :: Bugs]

#14892: DOTT: Crash when trying to play MM on Ed's computer
---------------------+----------------------------------
Reporter:  alexbevi  |       Owner:  (none)
    Type:  defect    |      Status:  new
Priority:  normal    |   Component:  Engine: SCUMM
 Version:            |  Resolution:
Keywords:            |        Game:  Day of the Tentacle
---------------------+----------------------------------
Comment (by eriktorbjorn):

 I was able to reproduce it with a scummvm.ini where Day of the Tentacle
 was the only game I had added.

 At this point, it's looping through all installed games trying to find one
 that's installed in a "maniac" sub-folder to the current one. One of the
 games it checks is Day of the Tentacle itself, i.e. it will check if a
 path is relative to itself.

 This leads to Path::extract("", nullptr) being called, where "" is one
 byte after what it calculates as the string's end point. Then it tries to
 create a new string from that begin/end, and that's where it crashes. I
 think.

 We could easily check that path != currentPath in the startManiac() loop,
 but it seems to me that path.relativeTo(currentPath) shouldn't be allowed
 to crash when path.isRelativeTo(currentPath) is true. Which leads to the
 question, should a path be considered relative to itself or not? Seems
 like a question for Le Philousophe.

 By the way, we could probably replace the
 path.toString('/').matchString("*maniac*", ...) check with something
 simpler now, after the recent "Migrate engine to Path" changes? Perhaps
 check (case insensitive) what path.getLastComponent() points to? It looks
 like that would strip away any trailing path separators, which was the
 point of using matchString() to begin with.
-- 
Ticket URL: <https://bugs.scummvm.org/ticket/14892#comment:9>
ScummVM :: Bugs <https://bugs.scummvm.org>
ScummVM