<Remi_M>
eregon: I don't remember if GCC 6.3 has all bugs fixed, but you can just disable some of its optimisations in case the result is crashing. It should be able to compile PyPy-STM. see the following description about what optimisations you may need to disable:
<eregon>
[platform:Error] implement_2.c:19015:34: warning: passing argument 1 of ‘ASN1_STRING_to_UTF8’ from incompatible pointer type [-Wincompatible-pointer-types]
<eregon>
[platform:Error] In file included from /usr/include/openssl/objects.h:965:0,
<eregon>
[platform:Error] from /usr/include/openssl/evp.h:94,
<eregon>
[platform:Error] from /usr/include/openssl/x509.h:73,
<eregon>
[platform:Error] from /usr/include/openssl/ssl.h:156,
<eregon>
[platform:Error] from common_header.h:127,
<eregon>
[platform:Error] from implement_2.c:4:
<eregon>
[platform:Error] /usr/include/openssl/asn1.h:1010:5: note: expected ‘unsigned char **’ but argument is of type ‘char **’
<eregon>
[platform:Error] int ASN1_STRING_to_UTF8(unsigned char **out, ASN1_STRING *in);
pilne has joined #pypy
kipras is now known as kipras`away
<Remi_M>
eregon: hm.. so yes, there seems to be a slight mismatch in the signature that we declare in the ssl module and your header file... If you just want a bare-bones pypy-stm, you can pass the --no-allworkingmodules option to targetpypystandalone
<LarstiQ>
eregon: yes, that's too recent
<LarstiQ>
eregon: openssl made a backwards incompatible change there
marky1991 has quit [Ping timeout: 248 seconds]
oberstet has joined #pypy
rokujyouhitoma has joined #pypy
<eregon>
ok, I found --withoutmod-_ssl hopefully that works :)
<eregon>
these options (like --no-allworkingmodules) have to be passed after targetpypystandalone.py, right?
rokujyouhitoma has quit [Ping timeout: 240 seconds]
<Remi_M>
eregon: yes exactly, I just hope nothing else depends on it :)
bendlas has quit [Ping timeout: 276 seconds]
nbjoerg has left #pypy [#pypy]
jcea has quit [Quit: jcea]
<eregon>
I think I have enough fractals to print myself a poster now :)
_whitelogger has joined #pypy
bendlas has joined #pypy
bendlas has quit [Read error: Connection reset by peer]
<palecsan>
Regarding my question earlier (about issue 2426), do you think is worth digging deeper?
Ryanar has quit [Quit: Ryanar]
bendlas has joined #pypy
rokujyouhitoma has joined #pypy
rokujyouhitoma has quit [Ping timeout: 240 seconds]
<ronan>
palecsan: well, it would be nice to solve it, but Armin's comments still apply AFAIK
nopf has quit [Remote host closed the connection]
<palecsan>
ronan: ok :) thanks
<palecsan>
if I want to specify an additional gcc flag (such as march) can I do this without modifying the generated C code in /tmp ?
<nimaje>
go to the dir in /tmp and modify the Makefile or use ´CFLAGS=... make´ :)
<palecsan>
there is no other way? :( when calling ../../rpython/bin/rpython?
<palecsan>
I think that also the small stub files must be generated/compiled with the same flags
<palecsan>
not just the testing_1 content
<palecsan>
or am I wrong?
<nimaje>
palecsan: if you call rpython you do a new translation, but yes I forgot to mention that you have to clean that dir from .o files, afaik you only need ´make´ in testing_1 (after cleaning and your changes to the Makefile)
<palecsan>
oh, and the make in testing_1 also re-builds everything in the parent directory?
<palecsan>
did not know that
<palecsan>
I assumed it's made only when translation is started
<palecsan>
thank you nimaje
<nimaje>
afaik, everything needed is in testing_1 (no idea why that dir has that name), but I had changed ´CC´ from ´gcc´ to ´cc´ (on a system where cc is clang) and removed something from LDFLAGS (as ld on that system didn't support that option and files compiled with it) and it worked
rokujyouhitoma has quit [Ping timeout: 248 seconds]
arigato has joined #pypy
Fleche_ has joined #pypy
Fleche has quit [Ping timeout: 240 seconds]
<kenaan>
tobweber stmgc[c8-fix-commit-signalling] 5d95b73a59c9 /c8/stm/: Signal commit to inevitable transaction if another transaction wants to commit
<kenaan>
tobweber stmgc[c8-fix-commit-signalling] 51b2dcee8206 /c8/stm/core.c: Add signalling also to become inevitable
<kenaan>
tobweber stmgc[c8-tcp-style-trx-length] 39f4ab9d3e31 /c8/stm/: Merge fix for commit signalling to inevitable transactions
<kenaan>
tobweber stmgc[c8-fix-commit-signalling] f27950321d3c /c8/stm/: Evaluate signal to commit only if running inevitable, and refactor wait implementation o...
<kenaan>
tobweber stmgc[c8-tcp-style-trx-length] e0f9b83d9771 /c8/stm/: Merge minor fixes to commit signalling
<kenaan>
tobweber stmgc[c8-adaptive-trx-length-per-thread] cf2b97b05bc6 /c8/stm/core.c: Merge add signal to commit inevitable trx when becoming inevitable
<StyXman>
I noticed that pypy's parser wants to keep the source in utf instead of unicode, even after detecting the encoding and converting from it. can someone explain the reason? I'm trying to port it to py3 to use it in a project of mine and this causes some problems...
<arigato>
I think that CPython's parser does the same, so we just copied CPython
<kenaan>
mattip cpyext-add_newdoc feb0a7a8e0bb /pypy/module/cpyext/test/test_typeobject.py: test for modifying tp_doc even after PyType_Ready
<kenaan>
mattip cpyext-add_newdoc 895424ab836c /pypy/: allow assignment to NULL tp_doc even after PyType_Ready
<kenaan>
mattip cpyext-add_newdoc 4e8cc878567f /pypy/: test, add more doc assignment from cpyext
mattip has joined #pypy
<mattip>
it seems we _can_ modify docstrings from c even after PyType_Ready,
<mattip>
perhaps the better question is if we _should_
<kenaan>
rlamy py3.5 a5510f559b24 /pypy/module/cpyext/: Move CPyBuffer and dependencies to buffer.py
<kenaan>
rlamy py3.5 73b9066a185c /pypy/module/cpyext/test/test_bytesobject.py: fix and update tests
<arigato>
mattip: imho, no
<arigato>
but if you feel like it, you can hack around in objspace/std/typeobject:descr__doc()
<mattip>
arigato: that is the point of the cpyext-add_newdoc branch, to support NumPy's add_docstring function
Taggnostr has quit [Ping timeout: 276 seconds]
<mattip>
with checks that the current __doc__ is None before modification
<mattip>
if we reject this approach, it means somehow rewriting numpy to add the docstrings before PyType_Ready
Taggnostr has joined #pypy
rokujyouhitoma has joined #pypy
<ronan>
mattip: I already implemented add_newdoc in RPython, perhaps you can reuse it?
<mattip>
ronan: I took a look, but it seemed more straighforward to do what I did in the branch
rokujyouhitoma has quit [Ping timeout: 268 seconds]