On Sat, 6 Dec 2003, Ferdi Franceschini wrote:

Hi Jorge,

On Fri, Dec 05, 2003 at 09:25:20AM -0300, Jorge Arellano Cid wrote:

...

Hi Ferdi,

I've catched an API problem inside the capi/cache that may be the cause for the fork-hog. This note is just to keep you from wasting time reviewing it. I'll write to you soon.

Great.

It _was_ the cause! There was a race condition hidden behind the lack of dpi-cache connection (the part we blamed) and a slightly broken API (the cloaked problem). As a matter of fact this bug had remained unnoticed maybe for years. In my patched tree (which is under test), the bug is fixed (both parts) and dillo works even better than before.

I think I have a new solution to the inf loop caused when a connection is not accepted on a dpi socket for any reason. This new solution does not require any code changes to the plugins. I will post more info to dillo-dev but I need to experiment a bit first.

That's good news. Cheers Jorge.-