Re: [SMARTY-DEV] smarty recompiling with op-code accelerators

Monte writes:
> That may help with Smarty upgrades, but does not address the original
> issue presented in this post. The accelerator op-code cache still won't
> refresh when the templates_c directory gets cleared (outside of a Smarty
> upgrade.)

The original issue was about deleting the cache "when upgrading Smarty or
making other major changes", so this would solve the problem in the first
case but not the second. In my recent experience the need to delete the
cache for anything other than a Smarty upgrade has been rare (I can't
remember a time when doing so other than when trying to determine the cause
of odd behaviour, under which circumstances I'd restart (or even disable)
the op-code cache anyway. In all cases where I've done this, emptying
templates_c hasn't fixed the problem anyway. Smarty is just too damn
reliable!

Therefore in my limited experience, including the version in the hash would
solve this problem well over half the time. (Thanks for mentioning it, btw:
it hadn't crossed my mind before.)

What do opcode accellerators normally check for? Just mtime? Does the inode
not come into the equation? (Of-course things are different with Windows
which also needs to be supported by any solution.)

IMHO we shouldn't really be telling the OS that the file hasn't changed when
it has, which appears to be what we're doing here - presumably that also has
implications for backups (if anyone is backing up templates_c, which I'm
sure some people are), and doubtless other situations which (like the
op-code cache) aren't obvious until they bite. Should we instead be looking
for a better way for Smarty to determine that the file matches the template
and doesn't need recompiling?

-- 
Mark Rogers,
More Solutions Ltd :: Tel: 0845 45 89 555