+ Standalone executable based on wheel.
+ Properly use `pkg_resources` so reading `welcome-msg` file
works even from within zip.
+ Mark `zip_safe=True` in setup.py.
+ Delete forgotten distutils startup script.
+ Build standalone before installing anything else,
to check if any deps are missing.
+ Restore py25 in Travis.
- Read welcome-msg in UTF-8.
- Add cmd-line option for `welcome-file`.
- Add TCs for welcome-file option and `/` http-req.
- Update docs for new option.
- Failback to in-code welcome-msg if unreadable (ie standalone mode, bad file).
* Use stabdard python's logging lib.
* Log http-request/response/errors.
* Cmd-line options for logging-format and filename.
* Cmd-line options for request /response/error requests/responses/errors
props to log.
* Add `-v` option controlling verbosity.
* Add docs about new options.
* TCs only `-v` & `--log_file ` (logging statements used throughout all
tests).
The fix in commit 7f97612 for supporting the package naming used by the
pytz module caused a regression if the package name contained a dash
followed by a number. We fix this by splitting on all dashes followed by
numbers and recreating the package name from all components but the
last.
The main line repo
(c1b9386dff)
currently lists duplicate entries on the "/simple" index page if the
package name contains hyphens and if both a source file and egg file are
uploaded. One index entry contains hyphens in the name which matches the
source file name and the other entry contains underscores in the name
which matches the egg file. Clicking both links lead to pages which list
the same files (both the source file and the egg file).
The previous fix
(c6966afeed)
mostly fixed this problem but it did not account for package names that
included an underscore. Change the "/simple" index to use the name from
a non-egg file, since this name should not have hyphens converted to
underscores as egg files do. If there are only egg files, use the egg
file name.
instead we use an iterator that returns all relevant information. that
simplifies the code quite a bit because we now mostly operate on the
iterators result instead of handling filenames, versions, prefixes...
I'd like to run multiple pypiserver wsgi apps inside one process and
plan to do so by reloading pypiserver._app.
this is the first step. we move all of the @route'd functions to _app.
