I’ve been working hard these days to release Reli 0.12.0, but it feels like it’ll never be finished. It’s that classic pattern where, just when I think I’m 90% done, the remaining 10% ends up taking up 90% of the total time.
- 7 Followers
- 44 Following
- 44 Posts
I usually play with projects like this. https://github.com/reliforp/reli-prof
@Crell Yeah, currently this is the first sentense I see.
> Earlier this week, Seifeddine Gmati, aka "ajezz," posted an RFC for PHP generics.
> Earlier this week, Seifeddine Gmati, aka "ajezz," posted an RFC for PHP generics.
@imsop I have to admit I was a bit worried for the last few days that my message on internals might have been completely off-base, or that I had phrased it badly, since it got no replies. https://externals.io/message/130816#130856
So seeing your latest idea around separating the runtime-checked layer from the SA-checked layer was honestly a relief. It seems close to what I was trying to get at, although your version generalizes it much better!
This is completely off-topic for the list, so I’m saying it here. :)
@shudder I guess I’m a vibe coder now 😇 but yeah, more output just means more details to deal with in the last 10%.
@Girgias And also spyish...? Sometimes I'm having difficulty distinguishing whether I'm making a profiler or a malware.
Some days ago I came up with an exciting hack to get the address of _tsrm_ls_cache for running ZTS threads from outside the process without an unstripped PHP binary (via brute forcing), but I haven't had time to try it yet...
The combo of php-memprof and Reli is golden. I'd like to add the ability to dump the addresses of allocated areas to php-memprof. Then combining both pieces of information together would give us a picture of which data occupy memory areas and when those areas were allocated.
As always, diagnosing memory_limit violations is fun.
https://github.com/vimeo/psalm/issues/10522#issuecomment-1881729504 #php #psalm
https://github.com/vimeo/psalm/issues/10522#issuecomment-1881729504 #php #psalm
A PoC for getting call traces and analyzing memory contents of scripts running on FrankenPHP by Reli is working on my end. But _tsrm_ls_cache works a bit differently than I imagined, so perhaps I don't understand ZTS or TLS handlings of ELF well enough.