Ok, landing this as-is!

Move CXXFLAGS into make command, per @jon's feedback.

Remove extraneous console.log(...) statement, per @atul's feedback.

Adding @atul and @jacek as blocking to get at least one more reviewer on this.

In D4725#136134, @ashoat wrote:

I agree! Although since the code for a lot of these diffs technically doesn't really depend on each other, I wasn't sure if to make a depends on.

Personally, I treat dependenices on Phabricator as a storytelling tool more than a way to actually indicate formal dependency relationships. Another way to think about it... a diff is a commit, and a stack is a branch.

In D4723#136124, @jon wrote:

@jon, is this something you'd like to do? I've made a Linear issue here. Feel free to cancel/backlog/whatever if it's not something we need to do.

I think we can do that conversation on linear. But the diff here should still go in

Update implementation of for loop. For some reason, the code committed was wrong and I changed it to include the variable/if check.

Fix line 18 -- was not adding the ABI to ABIS.

Taking a step back, this can eventually be replaced by just installing the service (e.g. make install), and running tunnelbroker or the related command without using this

Update line 12 of this diff. After reading through @ashoat and @karol's discussion on D2010 here, I realized why the regex was written like this. However, since ShellCheck complains about the right-hand side being quoted, I implemented @karol's other solution:

I could do a for loop here, but this is shorter and it works. If you think a loop would be better here, I can change it.

In D4718#136074, @varun wrote:

I think the real way to test that the script still works would be to make an invalid formatting change, like add some random line break in a rust file, and then try committing the change. If committing fails, the script still works as expected

I think you should be able to figure this out pretty easily by searching through the codebase

It would be more readable if you wrapped the diffs into a stack.

@karol, thanks for the response! Sorry, that is confusing. Let me clarify below and attach screenshots/output of the failures.

Adding @karol as blocking to get more context on the test plan.

Adding @karol as blocking for more context on the test plan.

In D4714#135639, @ashoat wrote:

I don't think we use this script anymore. I think it was only used in deploy.sh, which was deleted in D4484. I don't see any results with git grep setup.sh, anyways.

I think we can just delete it

Adding @atul and @tomek as blocking to get at least one more pair of eyes on this.

This script is part of the CMD layer, so it's what is going to be ran if we do run it, but we don't

Remove cut -d/ -f2- command because we want to keep the leading ./ on line 15

Exclude hidden files and directories using -not -path find flag instead of piping to grep

Nice!

In D4713#135601, @atul wrote:

Feel like there's probably a cleaner way to determine NEWEST_BACKUP, but this resolves the relevant shellcheck issue.

In D4716#135609, @atul wrote:

Outside of running shellcheck, we should ensure that the script continues to work as expected and specify how we can check that in the Test Plan.

eg something like:

build_openssl.sh is run during the Android build so if that workflow continues to pass CI we can reasonably assume that the script continues to work as expected