<phlebas> antocuni: i will check today, but twice is definitely not enough for us to warm up, we need more like 50 iterations 😉
<antocuni> phlebas: it's now "2 iterations", it's "2 runs of 100 iterations each"
<phlebas> i'm struggling with building the thing at all from the repo right now, I seem to be too stupid. It fails with "error: conflicting types for 'init_ujson_hpy_impl'" because we still define the init function with HPyContext rather than HPyContext* in the repo. I don't get why, though, that changed before 0.0.2 or am I completely confused?
<antocuni> phlebas: you need to checkout the hpy-0.0.2 branch of the ultrajson-hpy repo
<antocuni> the HPyContext vs HPyContext* thing changed between 0.0.2 and master, so it's HPyContext on 0.0.2 and HPyContext* on master
<phlebas> ah, i thought that was still in 0.0.2
<phlebas> is it expected that i cannot use tox with CPython in that hpy-0.0.2 branch? getting this:
<phlebas> ```
<phlebas> ```
<phlebas> Discarding git+https://github.com/hpyproject/hpy.git@release/0.0.2#egg=hpy.devel. Requested hpy from git+https://github.com/hpyproject/hpy.git@release/0.0.2#egg=hpy.devel has inconsistent name: filename has 'hpy-devel', but metadata has 'hpy'
<Hodgestar> antocuni: Running the "cindex.Index.create()" and then attempting to build the documentation segfaults for me, but documentation builds work fine without it. :/
<Hodgestar> antocuni: Does the same maybe happen for you on clang 10? (one needs a "make clean" first if one has built the documentation recently).
<phlebas> antocuni: locally I get
<phlebas> CPython: 180ms per iteration
<phlebas> Graal: 690ms per iteration.
<phlebas> But this is after 5 * 100 iterations, and we are still compiling even after 10 * 100 iterations, I haven't let it run further, but just to give you an idea - we have really, really long warmup.
jboi has quit [Quit: Bridge terminating on SIGTERM]
jevinskie[m] has quit [Quit: Bridge terminating on SIGTERM]
jboi has joined #hpy
jevinskie[m] has joined #hpy
<antocuni> phlebas: indeed, tox was broken on ultrajson-hpy@hpy-0.0.2. I just pushed a fix
<antocuni> Hodgestar: ouch indeed, it segfaults also for me :(
<antocuni> and also, sphinx_c_autodoc fails in my branch now:
<antocuni> docstring of runtime/argparse.c:1: WARNING: Error in declarator or parameters
<antocuni> Invalid C declaration: Expected identifier in nested name, got keyword: int [error at 17]
<antocuni> HPyAPI_HELPER int HPyArg_Parse(HPyContext * ctx , HPyTracker * ht , HPy * args , HPy_ssize_t nargs , const char * fmt , ...)
<antocuni> but it worked before with HPyAPI_RUNTIME_FUNC, so I don't know what's happening
<antocuni> maybe it no longer finds hpy.h?
<Hodgestar> antocuni: Re segfault: At least it is not just me. Perhaps we can just remove the "fix" again? :/
<Hodgestar> antocuni: Re build on your branch: It works for me on the branch I made off of your header refactor one, but I did update the folders to search in `conf.py` to point to the right place, so I think you are correct that the only issue is that it is not currently finding `hpy.h` in your branch. Feel free to just steal the fix to conf.py from my branch.
<antocuni> what is your branch
<antocuni> ?
<antocuni> ah ok, found it
<Hodgestar> Woot. :)
<antocuni> I don't understand, though
<antocuni> I don't find any change in conf.py in your branch
<Hodgestar> Hmm. Maybe I didn't push it. :/
* Hodgestar checks.
<Hodgestar> antocuni: Apologies -- I didn't push it because I had edited conf.py locally to work around the segfault. Pushed now.
<Hodgestar> I also made the formatting nicer. :D
<antocuni> thanks
mattip has quit [Ping timeout: 255 seconds]