BF: fix execution of patool commands

[yarikoptic]

It is an incomplete fix, which would still cause test failure in the crawler if no p7zip is installed in crawler due to #3175 , but wanted to see if anything breaks here before proceeding further

[codecov]

Codecov Report

Merging #3176 into 0.11.x will decrease coverage by 32.82%.
The diff coverage is n/a.

@@             Coverage Diff             @@
##           0.11.x    #3176       +/-   ##
===========================================
- Coverage   90.83%   58.01%   -32.83%     
===========================================
  Files         249       92      -157     
  Lines       32783    14652    -18131     
===========================================
- Hits        29780     8500    -21280     
- Misses       3003     6152     +3149

Impacted Files	Coverage Δ
datalad/plugin/addurls.py	18.32% <0%> (-81.37%)	⬇️
datalad/support/globbedpaths.py	21.25% <0%> (-78.75%)	⬇️
datalad/interface/rerun.py	18.48% <0%> (-77.73%)	⬇️
datalad/interface/run.py	24% <0%> (-76%)	⬇️
datalad/plugin/export_archive.py	24.65% <0%> (-75.35%)	⬇️
datalad/interface/add_archive_content.py	19.23% <0%> (-71.16%)	⬇️
datalad/plugin/add_readme.py	29.31% <0%> (-63.8%)	⬇️
datalad/plugin/wtf.py	19.17% <0%> (-62.56%)	⬇️
datalad/metadata/metadata.py	23.39% <0%> (-62.12%)	⬇️
datalad/metadata/aggregate.py	14.24% <0%> (-60.24%)	⬇️
... and 212 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 9059059...06ab803. Read the comment docs.

[yarikoptic]

grr... Not sure which way to proceed here. So

• so far support for .gz relied pretty much on the uniform implementation across all archives, where we just extract an archive under a directory and look at the files extracted
• so it was inline with how p7zip handles all of them including .gz and p7zip does use stored in the .gz header original filename
• not sure yet what is the behavior if no filename was stored in the header

I thought that may be we could provide our custom gunzip functionality which would respect that filename and thus would mitigate discrepancy, but

• gzip/gunzip itself does not care about possibly stored original filename in the gzip header (unless given explicitly --name option)
• Python's gzip library explicitly reads and discards (does not even store anywhere, bleh) possibly present in the header filename: https://github.com/python/cpython/blob/master/Lib/gzip.py#L427

so I do not see an easy way to figure out that filename in the header. But we can't easily switch to "just use original filename stripped of .gz" since

• we might be operating already on a annex key (so no filename pointing to it)
• would require implementing .gz-special handling of datalad-archive urls. It would actually be the only proper solution probably since that stored filename is not really something to rely on (later on could be extended to other .bz2, .xz etc; I don't think xz even stores the original filename). But once again -- currently datalad-archive relies on the handling for those as for any other archive. So we had urls like dl+archive:MD5E-s24--5590a3cbf10c77d61fb4058b741760d9.gz#path=123&size=0 not dl+archive:MD5E-s24--5590a3cbf10c77d61fb4058b741760d9.gz#path=MD5E-s24--5590a3cbf10c77d61fb4058b741760d9&size=0 or even more adequate here just dl+archive:MD5E-s24--5590a3cbf10c77d61fb4058b741760d9.gz#size=0 (without any path within)
• so if we were to provide "proper" treatment/url we might need to provide compatibility layer just in case there is some datalad dataset with pointers to previously .gz'ed files. I.e. if there is a path for any of those .gz, we pretty much ignore it and verify that only a single file is extracted and take it.

[yarikoptic]

ok -- for now just decided to blow in no 7z is installed.

[yarikoptic]

oops -- forgot about our bright future, changed base to be 0.11.x

[kyleam]

Thanks for summarizing the results of all your digging.

ok -- for now just decided to blow in no 7z is installed.

Sounds sensible.

[yarikoptic]

with @kyleam 's blessing will merge for now

[yarikoptic]

Also we might want to consider moving this into the runner itself. I didn't spot any place where we use shell and pass the list, so it shouldn't change any other behavior but might prevent us from hitting it again