[#132] Base py3 migration. #676
Conversation
660f0de to
6cda862
Compare
32ad757 to
b8277fb
Compare
b8277fb to
0acc6f8
Compare
|
From Mișu i guess the ones with glibc older than 2.26 should be removed when switching to a newer py3 version: 10:13 amazon 2018.03, centos 5/6/7, ubuntu 14.04/16.04 |
b9ab562 to
d4cbe39
Compare
d4cbe39 to
ddd6aa3
Compare
b2b9065 to
a56c5e6
Compare
|
python3 bare tests are green... I guess we can switch to 3.11 :) |
|
I am going to merge this... the future is py3. We can create a separate branch is we need py2 backward compatibility releases |
|
Dan, please take a quick look at this. The main changes are tested in chevah/server but we also need a good test coverage here. note that this is the initial py3.8 preparation work. The final green test run is with python 3.11 at #693 needs-review |
| ; Required for some unicode handling. | ||
| unidecode | ||
|
|
||
| wmi ~= 1.5.1; platform_system == "Windows" |
There was a problem hiding this comment.
; platform_system == "Windows"
Is this comment also code?
There was a problem hiding this comment.
yes... this is code...
https://setuptools.pypa.io/en/latest/userguide/dependency_management.html#platform-specific-dependencies
I guess that we should migrate to .toml files ... but in a separate PR.
| @@ -763,6 +756,11 @@ def _dirEntryToFileAttributes(self, entry): | |||
| if path.startswith('\\\\?\\') and path[5] == ':': | |||
| path = path[4:] | |||
|
|
|||
| hardlinks = stats.st_nlink | |||
| if not hardlinks and os.name == 'nt': | |||
| # I don't know why on Windows we doing scandir. | |||
There was a problem hiding this comment.
Maybe speed?
Using scandir() increases the speed of os.walk() by 2-20 times (depending on the platform and file system) by avoiding unnecessary calls to os.stat() in most cases.
From the package that is now included in Py3.
There was a problem hiding this comment.
The comment was bad... the comment was about scandir returning hardlinks=0
we use scandir for speed and most importantly not to block the process.
With os.listdir, if you have 100.000 files in a dir, when you call it it will return all of them in one go, possible blocking the process.
with scandir we have an iterator
| # On Windows the folder will have a size. | ||
| # So we skip this check on Windows. | ||
| result = self.filesystem.getFileSize(segments) | ||
| self.assertEqual(0, result) |
There was a problem hiding this comment.
Would it be any use to check the actual size of the file linked, in the case of Windows?
src/chevah_compat/unix_filesystem.py
Outdated
| if not isinstance(tail, text_type): | ||
| tail = tail.decode('utf-8') | ||
| if not isinstance(tail, str): | ||
| tail = tail |
Scope
Fix #132
This is the migration of the compat code to py3.
It removes py2 support as we don't want to look back :)
Changes
NT and Unix services compat were removed as they are no longer used.
We now use external tools to manage our daemons.
NT command line unicode converting code was removed as it's no longer an issue with py3.
The package definition was "modernized" to declare everyting inside setup.cfg
the package was moved outside of the namespace package.
With a new package name, in theory we can have both old and new compat installed at the same time.
it was also moved into a dedicates 'src' path.
The change was easy as we have already prepared most of the compat code for py3.
How to try and test the changes
reviewers: @danuker
check that changes make sense.
We will know for sure if this work, once we integrated it with chevah-server