Re: simple valgrind test found what appear to be serious memory problems

() [email protected] (Karl Berry)
() Thu, 30 Sep 2010 20:56:43 GMT

   One approach that would cover a lot of ground is a comprehensive
   test suite.

Good idea.

The infrastructure is there in subdir tests/ (in repo, renamed from
v/ in released tarball), and centralized (commit 8ea2149) for easier
wrangling.  I think i will formalize the test number scheme, moving
t00[12] to some higher range and reserving the numbers 000-?ff for
basic functionality, to be filled in along the way.

   Another idea would be to forget the complete rewrite for a time
   and do a "minimal" update of the infrastructure files, switching
   to GPLv3, and other such non-invasive changes.

Intertwined with the "many many implementation details" changes are
the "many many build methods" (i.e., autoconf / automake / gnulib)
changes.  That was a slog i wouldn't relish retracing later.

Also, i don't think "complete rewrite" is accurate.  I left time/date
handling pretty much intact (pending gnulib settling) as well as core
editing routines, use of global vars in the drivers (foo.c w/ ‘main’),
etc.

   P.S. I recommend staying away from obstacks.

OK, i'll keep this in mind for future (mis)doings.