[cmd diff] Run names with literal : in them understood differently than store and cmd runs commands

[whisperity]

Describe the bug

If the run name is in some way "complex", cmd diff constantly reports No run names with the given pattern:. However, the exact same format of run name is accepted by cmd results (and also store).

The complex run-name in question looks like this: my/complex-run: name. (Auto-generated from Git user/repository: branch format.)

CodeChecker version

612b726

To Reproduce

1. store --name simple-1
2. store --name simple-2
3. store --name "my/complex-run: name"
4. cmd results "my/complex-run: name" works. ✔️
5. cmd diff --new --basename "simple-1" --newname "simple-2" works. ✔️
6. cmd diff --new --basename "my/complex-run: name" --newname "simple-1" doesn't work. ❌

Expected behaviour

The run name is accepted, and before pattern-matching(?) or file system traversal, exact string match is done for the remote name. (In addition, if there are BOTH a local directory structure AND a remote run with the same pattern(?), a warning is emitted.)

System

• OS: Ubuntu 18.04

Additional context

The issue appears both when diffing remote to local and remote to remote, and in both directions.

If I make a directory locally, and execute diff, it will pick the directory up as the local run. That is, if the following directory structure appears locally (even if empty) ...

$ tree my
my
└── complex-run: name

... then all the remote findings will be regarded as --new, and reported.

I tried escaping most of the suspicious characters in the run name, i.e. "my\/complex-run: name", "my\/complex\-run\:\ name", and various other combinations of escape, but had no luck.

The exact error message is a bit misleading, I had to have the hunch to try making a directory, because the following relevant lines are printed, even with --verbose debug.

[INFO][2021-12-06 11:25:31] {system} [14277] <140079117096768> - cmd_line_client.py:1190 handle_diff_results() - Given remote runs (--baseline): my/complex-run: name
[DEBUG][2021-12-06 11:25:31] {system} [14277] <140079117096768> - client.py:257 setup_client() - Initializing client connecting to [...] done.
[WARNING][2021-12-06 11:25:32] {system} [14277] <140079117096768> - cmd_line_client.py:156 process_run_args() - No run names match the given pattern: my/complex-run: name

It does not even say that it is trying to find local directory structures.

[csordasmarton]

@whisperity The client thinks that you gave complex-run as the run name and name as the history tag name:

codechecker/web/client/codechecker_client/cmd_line_client.py

Line 149 in 612b726

run_with_tag = run_arg_with_tag.split(':')

[whisperity]

The client thinks that you gave complex-run as the run name and name as the history tag name

@csordasmarton Ah, thank you. Then it seems local directories do indeed take priority!

Would it be possible to implement some sort of an "escape" mechanism here, e.g. first match for \:, then do not split, and rewrite as :? The issue is that other commands happily accept : as-is, without trying to understand it as a tag. (store has a dedicated --run-tag or somesuch CLI option, I'm not sure about cmd results.)

Or should I just come up with a better run name generating logic, and add something to my documentation to discourage users from putting : into their run names (if they configure it themselves)?

[csordasmarton]

I think we can support escaping here, similar to this:

>>> run_name="my/comple-run\: name:tag_version"
>>> list(map(lambda x: x.replace("\:", ":"), re.split(r"(?<!\\):", run_name)))
['my/comple-run: name', 'tag_version']

[whisperity]

@csordasmarton Perfect! 😃 Will you have some time to add this small feature, or should I?

[csordasmarton]

If you have time, please create a PR for it. Also don't forget to add a test case for it 😊