Re: [Gnu-arch-users] TLA commit

[John A Meinel]

Ron Parker wrote:

> I am just curious.  Has anyone ever had "tla commit" go off into never
> never land and not complete?
>
> I am trying to commit a tree of Apple specific changes to a gcc
> archive I have created and it has been running for +10 hours on my
> Powerbook.  Currently the ,,commit... dir is about 648MB and tla's
> real memory usage is ~80MB.
>
> I'm running tla 1.3.1fix1.
I haven't seen that, but I do know of some specific memory corruption
that exists in tla-1.3.1. Specifically with Tom's new rel_table
implementation, he doesn't always initialize his pointers to 0. I know
of a specific location where this causes random memory to be overwritten
(during the delete function).

Checkout my patch
address@hidden/tla--pcomp--1.3.1--patch-12

This goes through and makes sure all rel_field entries are properly null
terminated.

Just for reference, the specific corruption that I found was in
changeset-report.c line 301. If you have a changeset which modifies a
files permission, then random memory gets corrupted because id_field
never gets set.

John
=:->

signature.asc
Description: OpenPGP digital signature