Robert Spier wrote:
It's working fine for a lot of folks without the patch.
I've never seen that behavior myself (though I am hardly the highest volume forkserver user). Personally, I suspect that it is completely load dependent, and relates to the not-all-that-great signal behavior than an interpreted language like Perl is always going to struggle with. Perl is fat and slow (in the best possible way ;-), so overlapping signals are going to be a problem forever.
One big negative though - your patch adds tabs to a file that uses space indentation. Please be consistent.
Never fear, that is a personal bugaboo of mine so I always scan patches before committing for just that sort of thing (I also deal harshly with >80 character lines). Maybe after 0.30 is out the door, I'll make a branch and apply perltidy to all the source files so people can see what *that* would look like.
As for the rest of the patch, there's a potential downside which is that children will hang around until the next connection comes in.
On this point, I think that I agree with Peter in that the only people who might have this be a problem are likely to be exactly the sites where the load will be so high it won't be a problem at all.
John
