ci: Use optimized Debug build type in test-each-commit

maflcko commented at 8:25 am on July 7, 2025: member

An optimized debug build is mostly as fast as a release build, because hot loops of heavy debug-only code are rare. So use that setting in the test-each-commit CI, to enable more checks almost “for free”.

DrahtBot added the label Tests on Jul 7, 2025

DrahtBot commented at 8:25 am on July 7, 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/32888.

Reviews

See the guideline for information on the review process.

Type	Reviewers
ACK	Prabhat1308, willcl-ark

If your review is incorrectly listed, please react with 👎 to this comment and the bot will ignore it on the next update.

Conflicts

Reviewers, this pull request conflicts with the following ones:

#31802 (Add bitcoin-{node,gui} to release binaries for IPC by Sjors)

If you consider this pull request important, please also help to review the conflicting pull requests. Ideally, start with the one that should be merged first.

maflcko force-pushed on Jul 7, 2025

Prabhat1308 commented at 12:04 pm on July 7, 2025: contributor

ACK fa21d3c

changes seem good as discussed in #32829 . Can we use mold as a linker in other Linux based system workflows ? dependencies we have seem to satisfy the deps here https://github.com/rui314/mold?tab=readme-ov-file#how-to-build

maflcko commented at 12:12 pm on July 7, 2025: member

Can we use mold as a linker in other Linux based system workflows ?

Sure, happy to review a follow-up. Only place to avoid it would probably the ci tasks that mirror the guix build (win-cross, mac-cross)

theuni commented at 2:43 pm on July 7, 2025: member

Is it necessary to use -g here? The output in #32829, for ex, doesn’t show a backtrace. So at least in that case the debug symbols are just taking up space.

maflcko commented at 2:50 pm on July 7, 2025: member

No, -g isn’t needed for the CI itself. It is only there for developers who happen to run this locally in a container and want to reproduce something in a debugger without having to modify the script and re-compile. Storage is cheap/free, so I thought it can’t hurt to have it, but no strong opinion. Anything should be fine for the setting.

maflcko commented at 8:51 am on July 8, 2025: member

To provide some numbers that the new checks really are roughly for free:

This ci run took 30min7sec (https://github.com/bitcoin/bitcoin/actions/runs/16112781421/job/45459901801?pr=32888)

Previously (randomly last month) it took 30min14sec (https://github.com/bitcoin/bitcoin/pull/32680/checks)

DrahtBot added the label Needs rebase on Jul 11, 2025

ci: [doc] reword debug log message fa21c3401e

ci: Use optimized Debug build type in test-each-commit faa3171ff2

maflcko force-pushed on Jul 14, 2025

Prabhat1308 commented at 7:46 am on July 14, 2025: contributor

re-ACK faa3171

DrahtBot removed the label Needs rebase on Jul 14, 2025

maflcko requested review from willcl-ark on Jul 18, 2025

in .github/ci-test-each-commit-exec.py:38 in faa3171ff2

35+        # Use mold, because it is faster than the default linker
36+        "-DCMAKE_EXE_LINKER_FLAGS=-fuse-ld=mold",
37+        # Use Debug build type for more debug checks, but enable optimizations
38+        "-DAPPEND_CXXFLAGS='-O3 -g2'",
39+        "-DAPPEND_CFLAGS='-O3 -g2'",
40+        "-DCMAKE_BUILD_TYPE=Debug",

willcl-ark commented at 2:24 pm on July 18, 2025:

Is -03 chosen here to try and counteract the slow-down from Debug as opposed to using -02?

IIRC I did some tests of -02 vs -03 and didn’t find much/any speedup, so just curious if that was the reason. If it’s been chosen to expose new classes of bugs not found by -03, then seems fine to me.

How does -03 actually play with -g2? Do the agressive optimisations play OK with debugging? (I’ve never tried this combination I don’t think.)

maflcko commented at 2:31 pm on July 18, 2025:

Probably doesn’t matter much if this is O2 or O3. Maybe O3 is 1% - 4% faster, according to #31491 (comment) ? Any of those can probably be changed, if there is a reason or need.

The main goal here is to add Debug (and the checks it ships with)

willcl-ark approved

willcl-ark commented at 2:25 pm on July 18, 2025: member

crACK faa3171ff22fea1c001e5a9b01d964aa425a3387

We can’t see this in action in the CI here as test-each-commit job omits the top commit (with the meaningful changes in in this case), but I did run locally and tested that the flags were being correctly appended.

Left a few questions regarding the flag selections, but ACK running these checks in this job.

maflcko requested review from fanquake on Jul 24, 2025

in .github/ci-test-each-commit-exec.py:34 in faa3171ff2

31         "-Werror=dev",
32+        # Use clang++, because it is a bit faster and uses less memory than g++
33         "-DCMAKE_C_COMPILER=clang",
34         "-DCMAKE_CXX_COMPILER=clang++",
35+        # Use mold, because it is faster than the default linker
36+        "-DCMAKE_EXE_LINKER_FLAGS=-fuse-ld=mold",

fanquake commented at 11:32 am on July 24, 2025:

The CMake here is one version too old, otherwise we could also have used CMAKE_LINKER_TYPE=MOLD

fanquake commented at 11:43 am on July 24, 2025: member

Seems fine to add this and be using mold in one job.

This ci run took 30min7sec (https://github.com/bitcoin/bitcoin/actions/runs/16112781421/job/45459901801?pr=32888)

This doesn’t actually show the new changes here right?

fanquake merged this on Jul 24, 2025

fanquake closed this on Jul 24, 2025

maflcko deleted the branch on Jul 24, 2025

fanquake commented at 1:23 pm on July 24, 2025: member

A run of the new job / flags here: https://github.com/bitcoin/bitcoin/actions/runs/16495974354/job/46641652091?pr=33044.

ci: Use optimized Debug build type in test-each-commit #32888

Code Coverage & Benchmarks

Reviews

Conflicts