Another bug

Robert Graham Merkel
Mon, 6 Nov 2000 13:22:50 +1100

Rob Browning writes:
 > Robert Graham Merkel <> writes:
 > > No, nothing like that.  I've traced the crash with gdb, and it
 > > happens in g-wrap generated code calling a C string->scm type
 > > conversion function with what appears to be a legitimate C string.
 > > Odd.
 > Any chance this is a "running out of RAM" bug?  I know there have been
 > a lot of g-wrap induced memory leaks, some of which the recent
 > const-string fixes should eliminate, but I seem to recall thinking
 > that we need to make sure we're using cleanup/no-cleanup when we're
 > supposed to...
It doesn't appear to be a RAM problem (top shows that the memory used
stays static), and the symptoms vary according to the guile version
used.  It appears to be a problem with the design of the transaction
report's HTML generator functions (they cause very large numbers of
string allocations and deallocations, which overloads the garbage
collector).  At least, that's Dave's theory, and I think he's right.
Anyway, it's not something that can be easily fixed - the best thing
to do is fix the cause by redesigning our HTML generator functions :)

BTW, I realise that my gtkhtml is an older version, and checking their
CVS it appears that page splitting has been improved in more recent
code.  I'll have another play with it now.

Robert Merkel