Re: FYI, non-locked cache value being modified

On Oct 24, 2006, at 11:48 PM, Gopal V wrote:

(I didn't actually get this mail ... but first things first) That code is a known breakage. Sara made the include_once_override as disabled by default in the last release. I haven't meddled in the optimization area because Ilia is rewriting it anyway (hint: run without it till then). And one of the intentions of putting the deep_copy flag in the op_array reserved section was to let such code figure out whether the op_array is in shm or local mem, to do such "stuff".

Right, good points. I guess I was curious how and if it's something that should be fixed, and the deep_copy flag answers half that question (sorry for not realizing it).

If you want to debug this further, I have a patched valgrind which lets me log memory writes [1].

Thanks for the patch, I'll play around with it...

Cheers, Gopal [1] - http://t3.dotgnu.info/code/valgrind-watchpoints.patch --The difference between insanity and genius is measured by success