icu-generic.gyp refactor - merge similar actions

[refack]

Spin off of #11217

After we exorcise the voodoo, it's time to merge some very similar duplicated code
(After @bnoordhuis's observation, the trim and code-gen actions were parameterized and merged)

Checklist

• make -j4 test (UNIX), or vcbuild test (Windows) passes
• commit message follows [commit guidelines][]

Affected core subsystem(s)

build, intl

[refack]

@nodejs/collaborators whomever can CI this, please. I think is finally ready.

[Trott]

CI: https://ci.nodejs.org/job/node-test-pull-request/7243/

[refack]

Refactored parameter selection completely
Ready for another CI...

[Trott]

CI: https://ci.nodejs.org/job/node-test-pull-request/7259/

[bnoordhuis]

This looks much more complicated than I would have thought necessary. Why is that?

[bnoordhuis]

This introduces a lot of churn for no good reason. Please don't do that, no reviewer likes a noisy diff.

[refack]

😛😛😛😛😛😛😛😛😛😛😛😛

[refack]

Sorry about that, it was the "old me", that tried to be cheeky.
Anyway I reverted, as requested.

[refack]

CI: https://ci.nodejs.org/job/node-test-pull-request/7259/

YAAAAAASSSSSSSS

[refack]

'icu_data_file' is not used anywhere anymore so I ☠️ed it

[refack]

Well it's back

[refack]

@bnoordhuis reduced churn (although I personally like "churn" if it's towards a cleaner code)

This looks much more complicated than I would have thought necessary. Why is that?

I took me too long because I assumed the 4 paths (win +/- small +/-) had more in common then what turned out. I now assume they just diverged over the years. For instance we need to repack for endiness only on non-windows. and genccode can only output .c files on non-windows, while it can directly compile to an .o file on windows.
IMHO the simplest/robust way was to parameterize the tasks, and set the parameters in a 4 clause condition. That way I was able to make the output to be as close as I could to the "baseline"

[refack]

Also

'action': [ 'python', 'icutrim.py', '-P', '<(PRODUCT_DIR)/.', # '.' suffix is a workaround against GYP assumptions :('-D', '<(icu_data_in)', '-T', '<(icudata_trim_temp)',
'--delete-tmp', ...

took some time to figure out since

1. I had to fight GYP not to make a mess of arguments that end with \
2. Then I realized --delete-tmp deletes the whole content of the directory -T points to before it operates, so we need a special <(icudata_trim_temp) directory and then copy the files back to where other actions can see them.
3. Also that the .gyp file is in tools/icu while the code is in deps/icu-small makes the whole thing cumbersome for a human to reason about.

P.S. I can put some of these comment it the code. Have put them in...

[gibfahn]

(although I personally like "churn" if it's towards a cleaner code)

Me too. If you're doing that you can do it as a separate commit (called squash! refactor code for readability or something), then reviewers can look at the other changes separately, and it can be squashed before merging.

[refack]

(called squash! refactor code for readability or something)

Good tip. Next time.

[srl295]

Nononono. Icu_data_file is used client side!! El abr. 7, 2017 4:02 AM -0700, Refael Ackermann <notifications@github.com>, escribió:

…

> > (called squash! refactor code for readability or something) > Good tip. Next time. — You are receiving this because you are on a team that was mentioned. Reply to this email directly, view it on GitHub (#12218 (comment)), or mute the thread (https://github.com/notifications/unsubscribe-auth/AA0Ms0Qfn7f7LiepCMZx7GIiofq_Sbflks5rthe7gaJpZM4MzT6O).

[refack]

Nononono. Icu_data_file is used client side!!

who is the client?

[refack]

Nononono. Icu_data_file is used client side!!

@srl295 like native addons?
If it's used anywhere I'll use it in icu-generic.gyp to actually set the name of the output 🤔...

[srl295]

Add on packages.. use it to find the file. Please leave the variables in the hash.. sorry it's complex we can hash it out if you want. Thanks for trying to simplify El abr. 8, 2017 11:33 AM -0700, Refael Ackermann <notifications@github.com>, escribió:

…

> > Nononono. Icu_data_file is used client side!! > @srl295 (https://github.com/srl295) like native addons? If it's used anywhere I'll use it in icu-generic.gyp to actually set the name of the output 🤔... — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub (#12218 (comment)), or mute the thread (https://github.com/notifications/unsubscribe-auth/AA0Ms2LBgfWXbWGCECr61gF7cesWfVPVks5rt9MNgaJpZM4MzT6O).

[refack]

Add on packages.. use it to find the file. Please leave the variables in the hash.. sorry it's complex we can hash it out if you want. Thanks for trying to simplify

Got you! so I'll use it in icu-generic.gyp to actually set the name of the output 👍

[srl295]

Please accept my +1 for the sentiment. Real review when I get a chance. I'd best explain it all over $Beverage .. El abr. 8, 2017 3:46 PM -0700, Refael Ackermann <notifications@github.com>, escribió:

…

@refack commented on this pull request. In configure (#12218 (comment)): > # this is the input '.dat' file to use .. icudt*.dat # may be little-endian if from a icu-project.org tarball - o['variables']['icu_data_in'] = icu_data_in - # this is the icudt*.dat file which node will be using (platform endianness) - o['variables']['icu_data_file'] = icu_data_file Well it's back — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub (#12218 (comment)), or mute the thread (https://github.com/notifications/unsubscribe-auth/AA0Ms55NAcWqykKPSCemEoX-7qHKXHd4ks5ruA5AgaJpZM4MzT6O).

[refack]

@srl295 tell me more about it, I don't get it. That file is not vendored by node-gyp for example...

[bnoordhuis]

@refack Are you still working on this?

[refack]

@bnoordhuis It's stable.
I tried to address your comment with some comments in the code.
IMHO this can land.
Next step is to do what you did in #12231 and move all the logic to a .py file.

[refack]

New CI: https://ci.nodejs.org/job/node-test-commit/9893/

P.S. since it's just a .gyp instead of trying to review the code, I can provide the before and after output variations (±windows × ±icu-small × make/msvs/ninja) for comparison ( react with 👍 If you want it )

[refack]

Cross-Ref: #7254

[refack]

@srl295 I see you're around.
Does it make sense to preprocess icudata and commit the processed files (trimmed/full * LE/BE)?

[BridgeAR]

Ping

[refack]

Closing in favor of a better idea.