Log file Viewer

[11:53:32] <peque> Hi there :-)

[11:53:35] <peque> I have uploaded the documentation of a project to pythonhosted.org (see https://pythonhosted.org/osbrain/). However, as you can see, the logo is not correctly displayed. If you "right click -> open", then you can see the logo is there. Also, when opening the generated documentation locally in my computer, the logo is correctly displayed. If I use a PNG logo instead of the SVG, then everything seems to work fine in pythonhosted.org. I

[11:53:37] <peque> s this normal? Can't I use a SVG logo with pythonhosted.org? why would this fail while the logo is actually there? Thanks in advance for your help :-)

[11:58:58] <peque> Also, when uploading the zipped HTML documentation through pypi, when a SVG is used, I get an error response from the server (although the documentation seems to be uploaded anyway afterwards): """Error 503 backend read error\n backend read error\n Guru Mediation:\n Details: cache-fra1235-FRA 1468409925 988906803"""

[13:08:03] <peque> btw, forget about the error when uploading the zipped HTML documentation. It seems it was just coincidence. I am now not getting that error, but still I have the same problem: SVG logo is not displayed when uploaded to pythonhosted.org

[14:28:45] <peque> I opened an issue in https://github.com/pypa/pypi-legacy/issues/480

[15:58:02] <nedbat> anyone know anything about this? https://bitbucket.org/ned/coveragepy/issues/504/42b1-overwrites-pythons-sitepy seems that installing coverage.py writes a site.py file that I didn't author? Not sure on what versions it happens.

[16:00:23] <tdsmith> apparently that file exists in the distribute source tree https://searchcode.com/codesearch/view/53542847/

[16:01:34] <tdsmith> also setuptools

[16:01:43] <tdsmith> that is all my googling skills have to contribute this morning

[16:11:45] <nedbat> tdsmith: thanks

[18:05:21] <wiggy> https://github.com/pypa/setuptools/blob/master/pkg_resources/__init__.py#L1373-L1378 and https://www.python.org/dev/peps/pep-0503/#normalized-names seem to conflict

[18:05:28] <wiggy> does anyone know which one is correct ?

[18:06:33] <apollo13> good question, if anyone knows, dstufft -- but since setuptools is used by pip, I guess whatever pip allows wins

[18:06:43] <apollo13> or rather the combination of pip + pypi

[18:08:17] <dstufft> wiggy: safe_name isn't for cannonicalization, it's for making an arbitrary name "safe" (although a fair number of folks have used it for canoncalization, the rules are almost the same, but not quite)

[18:08:33] <wiggy> dstufft: what is the right API for canonicalisation ?

[18:09:00] <dstufft> PEP 503, or https://packaging.pypa.io/en/latest/utils/#packaging.utils.canonicalize_name which is an implemtentationo f that

[18:10:54] <wiggy> thanks

[18:34:04] <ktdreyer> dstufft: regarding https://sourceforge.net/p/pypi/support-requests/632/ - would you please add my PyPI username (ktdreyer) as an administrator for the kobo package?

[19:12:39] <grant-h> I'm attempting to install sqlalchemy with pip on windows

[19:13:31] <grant-h> however the pip install does not succeed because it fails to install the package's c-extensions

[19:14:20] <grant-h> from what I can tell from researching this when a c-extension fails in a pip install it normally gives a notice and continues the install

[19:14:55] <grant-h> but in my case an error is thrown and the entire installation is stopped

[19:15:26] <grant-h> any ideas about how I could change this behavior?

[19:15:52] <grant-h> I'm using pip 8.1.2

[19:16:19] <grant-h> and original my python and pip are from miniconda

[19:16:29] <tdsmith> a minority of extensions have a pure-python backup if a c extension module fails to build

[19:16:35] <tdsmith> sqlalchemy is probably not one of them

[19:16:46] <tdsmith> in general that's fatal

[19:18:46] <grant-h> tdsmith: it looks like sqlalchemy can be built without the extension: http://stackoverflow.com/questions/20436528

[19:19:30] <tdsmith> ah, indeed

[19:19:42] <tdsmith> did you try the example they gave for building without the C extensions?

[19:19:43] <grant-h> this command does work for me `pip install --global-option='--without-cextensions' SQLAlchemy"`

[19:19:49] <tdsmith> perfect

[19:19:57] <dstufft> perhaps it's failing in a way that the setup.py doesn't detect as a compilation failure

[19:20:09] <tdsmith> that's not a convention and there's not a way to make pip do that for arbitrary packages

[19:20:43] <tdsmith> sqlalchemy's setup.py is apparently configured to accept the --without-cextensions parameter but that's specific to sqlalchemy

[19:23:06] <grant-h> tdsmith: ok, so generally speaking the fact that the install fails without that flag is expected behavior?

[19:23:46] <tdsmith> yeah

[19:24:09] <tdsmith> you could ask the sqlalchemy project to do something else like i think markupsafe does

[19:25:35] <dstufft> fwiw SQLalchemy does attempt to gracefully fall back to pure python

[19:25:40] <dstufft> when compilation fails

[19:26:44] <tdsmith> maybe it's broken. :D

[19:27:18] <dstufft> grant-h: can you pastebin the error?

[19:29:30] <grant-h> dstufft: here you go: https://gist.github.com/grant-humphries/f921a210b11cf3b6c69857d5955a6c71

[19:30:06] <dstufft> grant-h: ok yea, so the problem is just that the error it's failing with, isn't an "expected error" by SQLAlchemy for failing compilation

[19:30:12] <dstufft> you should report that to SQLAlchemy I think

[19:31:09] <grant-h> ok, thanks

[19:31:49] <grant-h> so just ticket it here: https://github.com/zzzeek/sqlalchemy, do you think?

[19:33:59] <dstufft> https://bitbucket.org/zzzeek/sqlalchemy/issues?status=new&status=open here I think

[19:37:44] <grant-h> dstufft: haha, that would make more sense given there are no public issues on the github page, thanks again for your help

[19:37:57] <grant-h> and tdsmith thanks to you also