Re: Fix to long-standing crashes in GC

[Luc Teirlinck]

Stefan Monnier wrote:

   When the GC seems to be involved what it typically means that there's
   a memory corruption somewhere.  The only reason why the GC gets involved is
   because the GC traverses all your memory, so it basically does (as a side
   effect) some kind of sanity check, so memory corruption is usually detected
   during GC.

I would guess that means that studying the backtrace would not be
terribly useful in this case, since the corruption could have happened
anywhere.  I know how to use GDB, but I do not have that much
experience debugging an aborting Emacs with GDB.  I have a live
process around.

The bug is of course not necessarily in Tramp, but could be in any of
the machinery used by Tramp.  I get crashes using Tramp with ssh the
entire time.  I have never observed a crash using an external transfer
method, like scp (although I tried), so en/decoding seems a likely
culprit.  I get strange error messages about invalid base64 data.

Sincerely,

Luc.