Add option to include whole repository #443
Conversation
- Added a section archiver with two options:
- method ('git'/'tar'): How to create the tar ball
- include_git (boolean): Whether to include .git folder
Signed-off-by: Cristian Le <cristian.le@mpsd.mpg.de>
|
This one seems to error when running |
|
Epel 7/8 seem to be using old python version. would really prefer the |
Signed-off-by: Cristian Le <cristian.le@mpsd.mpg.de>
|
Oh no, we have to make the code compatible all the way to python 2.7 😢 |
| if add_git_folder: | ||
| tar_append_cmd = f"tar --append -f {initial_tar} --transform 's|^{relative_git_dir}|{prefix}/|' {relative_git_dir}/.git" | ||
| run_command(tar_append_cmd) | ||
| elif method == 'tar': |
There was a problem hiding this comment.
I have some issues when using
[archiver]
method = tar
Traceback:
Traceback (most recent call last):
File "/home/jkadlcik/git/tito/src/tito/cli.py", line 919, in <module>
main()
File "/home/jkadlcik/git/tito/src/tito/cli.py", line 910, in main
CLI().main(sys.argv[1:])
File "/home/jkadlcik/git/tito/src/tito/cli.py", line 209, in main
return module.main(argv)
^^^^^^^^^^^^^^^^^
File "/home/jkadlcik/git/tito/src/tito/cli.py", line 392, in main
return builder.run(self.options)
^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/jkadlcik/git/tito/src/tito/builder/main.py", line 160, in run
self.tgz()
File "/home/jkadlcik/git/tito/src/tito/builder/main.py", line 577, in tgz
self._setup_sources()
File "/home/jkadlcik/git/tito/src/tito/builder/main.py", line 610, in _setup_sources
create_tgz(self.git_root, self.tgz_dir, self.git_commit_id,
File "/home/jkadlcik/git/tito/src/tito/common.py", line 895, in create_tgz
tarfixer.fix()
File "/home/jkadlcik/git/tito/src/tito/tar.py", line 325, in fix
self.process_chunk(chunk)
File "/home/jkadlcik/git/tito/src/tito/tar.py", line 292, in process_chunk
chunk_props[member] = int(chunk_props[member], 8)
^^^^^^^^^^^^^^^^^^^^^^^^^^^
ValueError: invalid literal for int() with base 8: ''
Reproducible when you run tito build --tgz within this project.
There was a problem hiding this comment.
Indeed, that's where I wanted some help because I don't understand what the tarfixer does and why it fails like that.
| @@ -427,6 +427,10 @@ def __init__(self, name=None, tag=None, build_dir=None, | |||
| self.tgz_dir = tgz_base | |||
| self.artifacts = [] | |||
|
|
|||
| # Details of how to make the tar repo | |||
There was a problem hiding this comment.
Can you please explain these options in man tito.props?
| @@ -427,6 +427,10 @@ def __init__(self, name=None, tag=None, build_dir=None, | |||
| self.tgz_dir = tgz_base | |||
| self.artifacts = [] | |||
|
|
|||
| # Details of how to make the tar repo | |||
| self.tgz_method = self.config.get('archiver', 'method', fallback='git') | |||
| self.tgz_include_git = self.config.getboolean('archiver', 'include_git', fallback=False) | |||
There was a problem hiding this comment.
I understand the situations when
- building a tarball from the last commit is useful
- building a tarball from the code as is, even with uncommitted changes, is useful
but I can't imagine a situation when using a tarball from the last commit but also containing the .git directory is needed. Do you have a use-case for this?
There was a problem hiding this comment.
Well, this was my problems when building scikit-build-core. It uses setuptools_scm that can detect the tags from the .git folder. They offer an alternative via .gitattributes and generating a .git_archival.txt file, but that also fails without #445. There could be similar projects, in which case that would be the easiest for prototyping or quickly setting up a build system.
Of course these would be discouraged for production releases.
Yeah, that's always fun 😄 But in your case it looks like an easyfix IIRC Python 2.7 doesn't have f-strings yet. |
Please take a look here |
Yeah, I've understood the basics that it is to fix the timestamps so that the tars are identical. But I don't quite understand why it would depend if the tar ball was created via Although, wouldn't this be irrelevant if we include the whole |
Still needs to be exposed to CLI as well. Please help me on that part.
Closes: #441