A bit of doc update, and HarBuzz built from source example #172
Conversation
|
I want to write a bit about virtualenv also, please don't pull yet. |
HinTak
force-pushed
the
master
branch
2 times, most recently
from
e7531ca
to
5944eb3
Compare
Final part of doc update. Closes pygobject#171
|
I can't seem to figure out how to do `pwd` (for the pythonpath part), but this is all I wanted to write. Please review and pull. |
| @@ -29,6 +29,9 @@ Requirements | |||
| Calling ``source ./tools/bootstrap.sh`` will put you in a | |||
| virtualenv with all dependencies installed (except graphviz). | |||
|
|
|||
| Afterwards, you may need to set ``PYTHONPATH`` to | |||
| ``_venv/lib/python*/site-packages/``, where ``*`` is the | |||
| version number of your python installation. | |||
There was a problem hiding this comment.
Something isn't right here, that should be handled by the virtualenv
There was a problem hiding this comment.
If I try to run ./tools/build.sh, it says it cannot find "pgi". _venv/bin/activate modifies PATH so there is a private copy of python, etc but still cannot find the libs. By the way python itself is small - there is libpython elsewhere.
I am not familiar with virtualenv, and documentation is poor. Is there any way of checkin what it is supposed to be doing and what it is not?
There was a problem hiding this comment.
I did a bit of digging, and found something curious. If I start python interactively, it can find pgi:
(_venv) [Hin-Tak@localhost pgi-docgen]$ python
Python 3.6.6 (default, Jul 19 2018, 14:25:17)
[GCC 8.1.1 20180712 (Red Hat 8.1.1-5)] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import pgi
>>>
and if I just run the content of pgidocgen.py it works:
$ python
Python 3.6.6 (default, Jul 19 2018, 14:25:17)
[GCC 8.1.1 20180712 (Red Hat 8.1.1-5)] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import sys
>>> from pgidocgen.main import main
>>> import pgi
>>> main(["pgidocgen.py","create","_docs", "HarfBuzz-0.0"])
but somehow running pgidocgen.py via tools/build.sh does not work.
| is in a ``src/.libs`` sub-directory. So you need to create a directory | ||
| ``gir-1.0`` further inside, copy ``HarfBuzz-0.0.gir`` over there, and set | ||
| ``$XDG_DATA_DIR`` to the **parent directory** of your newly created ``gir-1.0``, | ||
| ``$GI_TYPELIB_PATH`` to ``src``, and ``$LD_LIBRARY_PATH`` to ``src/.libs``. |
There was a problem hiding this comment.
an alternative is to install things to a prefix and point things to there, but I guess this will do.
There was a problem hiding this comment.
I also found that setting $XDG_DATA_DIR does not work; but it is looking for gir's in /home/Hin-Tak/.local/share/flatpak/exports/share/gir-1.0 which at least is one place under my control.
It is looking at gir's at only 4 places:
/home/Hin-Tak/.local/share/flatpak/exports/share/gir-1.0
/var/lib/flatpak/exports/share/gir-1.0
/usr/local/share/gir-1.0
/usr/share/gir-1.0
This suggests that setting prefix probably does not work either?
| # which is without cacert.pem . | ||
| if [ -f /etc/redhat-release ]; then | ||
| pip install --ignore-installed certifi | ||
| fi |
There was a problem hiding this comment.
I cannot quite figure out what the code is trying to do, but for each package, it tries to download something from somewhere? The doc for the cert package (from forensics.cert.org) says you should use the system's certificate collection instead, which is located under /etc/pki, I think.
|
somewhat related: I've filed #173 |
|
Actually pgi-docgen's installation was not not too bad. I was reluctant to run Running it was more an issue - as you saw, it took me a bit of time to figure out how to regenerate HarfBuzz's doc from dev source (after I made some gobject annotation changes); the cert package situation was curious though understandable - the packager wants you to use the code but not the bundled certificates (which can get out of date). I also was not expecting it to recurse through my system's GIR directory and regenerating dependent packages' docs when I only asked for one. As far as this pull goes, I think your only objection is the pythonpath one? I think it is not ideal either, and I like to figure out what's wrong if there is any tips in looking into that. |
|
Apologies, been a while. I'll come back to this and see if there is any more work to be done soon... |
The "gir-1.0" is not obvious - I needed to run strace to figure that out.