Re: [PATCH] Fix numeric overflow leading to failure to GC

On 29 Dec 2006, nix(a)esperi.org.uk said: No it isn't, the cursed thing has started leaking again. I hate intermittent bugs. However, the patch above definitely fixes a bug which definitely *can* cause leaks. It's just not the only leak. The leak I see now persists even if `object-memory-usage' is always low, but even given that it's odd to see this: buffer text markers extents other total ------------------------------------------------------ *nntpd* 3956004 312 0 1036 3957352 yet this buffer has no text in it! 3.9Mb of pure silicon vapour :) I'm back to investigating. First cut: instrumenting malloc() and free() so that I can verify that this is truly leaked and not a fragmentation problem, and seeing if it happens under --disable-debug (as --enable-debug changes object allocation in some fairly significant ways: if this remaining leak is debugging-only, that would explain why I didn't see it earlier when setting gc-cons-percentage to 0: I was running a --disable-debug build then.) _______________________________________________ XEmacs-Patches mailing list XEmacs-Patches(a)xemacs.org http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-patches