test: use notarized v28.2 binaries and fix macOS detection

Sjors commented at 11:15 am on July 8, 2025: member

Since #31407 macOS guix builds are signed and notarized. This was included in v29 and backported to 28.x.

This PR bumps the v28.0 previous release binary to v28.2 and adjusts the test that uses it. Additionally it no longer manually code signs binaries >= v28.2.

While testing on an M4 mac and redownloading all the binaries, I noticed that platform == "arm64-apple-darwin" doesn’t actually work. This initially used args.platform in #26694, but that was changed to just platform in #32219.

So the first commit switches this to use args.host. I manually tested on Intel macOS 13.7.6 that code-signing still isn’t needed there (when downloading using a script).

Also documented that you can set HOST.

DrahtBot added the label Tests on Jul 8, 2025

DrahtBot commented at 11:15 am on July 8, 2025: contributor

The following sections might be updated with supplementary metadata relevant to reviewers and maintainers.

Code Coverage & Benchmarks

For details see: https://corecheck.dev/bitcoin/bitcoin/pulls/32922.

Reviews

See the guideline for information on the review process. A summary of reviews will appear here.

LLM Linter (✨ experimental)

Possible typos and grammar issues:

mac -> Mac [proper noun capitalization for Apple’s Mac]

drahtbot_id_4_m

in test/get_previous_releases.py:214 in fbae1cf1ee outdated

210@@ -211,8 +211,9 @@ def download_binary(tag, args) -> int:
211 
212     Path(archive).unlink()
213 
214-    if tag >= "v23" and platform == "arm64-apple-darwin":
215-        # Starting with v23 there are arm64 binaries for ARM (e.g. M1, M2) macs, but they have to be signed to run
216+    if tag >= "v23" and tag < "v28.2" and platform == "arm64-apple-darwin":

maflcko commented at 11:37 am on July 8, 2025:

the code is still wrong. Usually, it doesn’t make sense to compare two values of completely unrelated types and a type-safe language would prevent this class of issues.

Also, have you tested this code branch?

Sjors commented at 11:52 am on July 8, 2025:

Oh oops, I did test it, but then broke it when switching the commit order.

Sjors force-pushed on Jul 8, 2025

Sjors commented at 11:59 am on July 8, 2025: member

cc @kdmukai, @m3dwards

test: fix macOS detection 5bd73d96a3

test: replace v28.0 with notarized v28.2

Since #31407 guix builds are signed and notarized. This was backported to v28, so bump the version.

c6dc2c29f8

in test/get_previous_releases.py:214 in 95ab4db5d7 outdated

210@@ -211,8 +211,9 @@ def download_binary(tag, args) -> int:
211 
212     Path(archive).unlink()
213 
214-    if tag >= "v23" and platform == "arm64-apple-darwin":
215-        # Starting with v23 there are arm64 binaries for ARM (e.g. M1, M2) macs, but they have to be signed to run
216+    if tag >= "v23" and tag < "v28.2" and platform.system().lower() == "darwin" and platform.machine() == "arm64":

maflcko commented at 2:20 pm on July 8, 2025:

Your new code still looks wrong, when someone sets HOST=x86_64-apple-darwin on an arm machine, where signing isn’t needed for that case.

Why not use args.host and keep the "arm64-apple-darwin" string?

Sjors commented at 2:31 pm on July 8, 2025:

I don’t understand the use case. Afaik you can’t run the functional tests on such a cross-compiled build?

maflcko commented at 3:00 pm on July 8, 2025:

I don’t have an apple, but i had the impression that rosetta will just run the x86_64 executable on the arm64?

Sjors commented at 3:09 pm on July 8, 2025:

Let me try if that actually works…

Sjors commented at 4:12 pm on July 8, 2025:

I switched to using args.host.

With the previous version:

0cd depends
1gmake HOST=x86_64-apple-darwin
2cd ..
3cmake -B build --toolchain /Users/sjors/dev/bitcoin-rosetta/depends/x86_64-apple-darwin/toolchain.cmake
4cmake --build build

The result is as expected:

0file -b build/bin/bitcoin
1Mach-O 64-bit executable x86_64

Meanwhile if you just call test/get_previous_releases.py, it’s going to download arm binaries:

0test/get_previous_releases.py
1file -b releases/v28.2/bin/bitcoind
2Mach-O 64-bit executable arm64

The functional tests pass just fine with this combo.

But using the undocumented HOST env it will get x86:

0HOST=x86_64-apple-darwin test/get_previous_releases.py

Which it will needlessly codesign, though it’s not harmful.

Sjors force-pushed on Jul 8, 2025

Sjors commented at 4:12 pm on July 8, 2025: member

I switched to args.host, see #32922 (review)

Also documented that you can set HOST.

in test/get_previous_releases.py:290 in 72c54492ea outdated

284@@ -284,7 +285,12 @@ def main(args) -> int:
285 
286 if __name__ == '__main__':
287     parser = argparse.ArgumentParser(
288-        formatter_class=argparse.ArgumentDefaultsHelpFormatter)
289+        formatter_class=argparse.ArgumentDefaultsHelpFormatter,
290+        epilog='''
291+        HOST can be set to any of the `host-platform-triplet`s triplets from

maflcko commented at 5:47 pm on July 8, 2025:

0        HOST can be set to any of the `host-platform-triplet`s from

maflcko approved

maflcko commented at 5:48 pm on July 8, 2025: member

lgtm

test: stop signing previous releases >= v28.2 609203d507

test: document HOST for get_previous_releases.py 4bb4c86599

in test/get_previous_releases.py:214 in 72c54492ea outdated

210@@ -211,8 +211,9 @@ def download_binary(tag, args) -> int:
211 
212     Path(archive).unlink()
213 
214-    if tag >= "v23" and platform == "arm64-apple-darwin":
215-        # Starting with v23 there are arm64 binaries for ARM (e.g. M1, M2) macs, but they have to be signed to run
216+    if tag >= "v23" and tag < "v28.2" and "arm64-apple-darwin":

maflcko commented at 5:51 pm on July 8, 2025:

I switched to args.host, see #32922 (review)

I don’t think you did. The line still looks wrong and is missing args.host.

Sjors commented at 6:25 pm on July 8, 2025:

Another rebase screwup, tried again… that should also fix the windows CI failure

Sjors force-pushed on Jul 8, 2025

test: use notarized v28.2 binaries and fix macOS detection #32922

Code Coverage & Benchmarks

Reviews

LLM Linter (✨ experimental)