Let push honor multiple publication dependencies declared via siblings

[mih]

The documentation for declaring publication dependencies allows for two different ways to declare multiple dependencies:

1. Declaring a single config item with a list of dependencies (space-separated)
2. Declaring a config item multiple times with a single dependency each.

Previously push() only supported (1), while siblings() produced (2).

This change lets push() handle both scenarios.

Fixes #6867

Thanks to @LoannPeurey for the report!

[codecov]

Codecov Report

Merging #6869 (ddf7c42) into maint (cefecae) will increase coverage by 0.89%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##            maint    #6869      +/-   ##
==========================================
+ Coverage   90.25%   91.15%   +0.89%     
==========================================
  Files         354      354              
  Lines       46113    46135      +22     
==========================================
+ Hits        41621    42053     +432     
+ Misses       4492     4082     -410     

Impacted Files	Coverage Δ
datalad/core/distributed/push.py	92.57% <100.00%> (+0.02%)	⬆️
datalad/core/distributed/tests/test_push.py	98.94% <100.00%> (+<0.01%)	⬆️
datalad/cmd.py	85.43% <0.00%> (-8.67%)	⬇️
datalad/log.py	82.86% <0.00%> (-5.58%)	⬇️
datalad/api.py	90.24% <0.00%> (-4.50%)	⬇️
datalad/customremotes/tests/test_datalad.py	95.34% <0.00%> (-2.28%)	⬇️
datalad/tests/test_log.py	98.61% <0.00%> (-0.70%)	⬇️
datalad/distribution/create_sibling.py	69.80% <0.00%> (-0.20%)	⬇️
datalad/cli/common_args.py	100.00% <0.00%> (ø)
datalad/tests/test_dochelpers.py	100.00% <0.00%> (ø)
... and 21 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update cefecae...ddf7c42. Read the comment docs.

[yarikoptic]

where do we document/allow for a white-space separated list of names?

Rationale for asking: Indeed ATM git does not allow white-space in the name of the remote, and it seems we do not "disallowed" it explicitly since just allow for git to error out:

$> datalad siblings add -s "my name" --url https://example.com
CommandError: 'git -c diff.ignoreSubmodules=none remote add 'my name' https://example.com' failed with exitcode 128
fatal: 'my name' is not a valid remote name

so if in the future git for some reason decides to support that (unlikely, but not impossible IMHO) such splitting by space might become buggy. So I wondered to ask where we already document/allow for that, and may be we should add some explicit checking for the name to not contain spaces (since could lead to fail of this added logic)

[mih]

Hmm, I cannot find a user-facing documentation that made me conclude that we allow(ed) for this. Maybe it was a comment, but I cannot find that either.

[yarikoptic]

then since we didn't have that documented, and looking at original code above and behavior of ensure_list:

$> python -c 'from datalad.utils import ensure_list; print(ensure_list("w1 w2"))'
['w1 w2']

that it doesn't split by space, I think we never had such a feature -- should we really add it along with this? may be just a matter of adding get_all=True and be done?

[yarikoptic]

@bpoldrack shines the archaeological light on the situation. We used to have get return all values by default, that caused issues, so get_all=False was introduced and thus broke prior behavior here which was to have all of the siblings pushed to (I guess the test didn't test for multiple). So I will push here a commit which replaces all the above with simple addition of get_all=True.

[yarikoptic]

2 travis fails known, one is something to note -- might need @slow

___________________ test_nested_pushclone_cycle_allplatforms ___________________
Test passed but took too long to run: Duration 30.47114537500005s > 30.0s
----------------------------- Captured stdout call -----------------------------

[yarikoptic]

with all of the above in mind, let's proceed !