[Backtracing] Add initial support for Windows. by al45tair · Pull Request #84906 · swiftlang/swift

al45tair · 2025-10-15T11:23:11Z

This PR includes support for capturing backtraces and symbolicating using DWARF debug information. It doesn't include support for PDB (yet).

rdar://101623384

al45tair · 2025-10-15T11:23:22Z

@swift-ci Please smoke test

al45tair · 2025-10-15T12:26:12Z

@swift-ci Please smoke test

al45tair · 2025-12-10T15:06:09Z

Status: this isn't working for 32-bit Windows, because of a problem with the Swift compiler itself.

Also, I need to work on the build system a bit because right now you need to manually copy the backtracer into the correct place. Not sure why, because I'm fairly certain there's CMake code already that's supposed to do that.

al45tair · 2026-01-12T11:07:35Z

@swift-ci Please test

al45tair · 2026-01-12T11:55:52Z

@swift-ci Please test

al45tair · 2026-01-13T12:25:31Z

@swift-ci Please test

al45tair · 2026-01-13T15:33:27Z

@swift-ci Please test

al45tair · 2026-01-14T11:17:39Z

@swift-ci Please test

al45tair · 2026-01-14T11:56:12Z

@swift-ci Please test macOS platform

al45tair · 2026-01-14T14:17:06Z

@swift-ci Please test

al45tair · 2026-01-14T15:36:09Z

@swift-ci Please test

al45tair · 2026-01-14T16:49:08Z

@swift-ci Please test

al45tair · 2026-01-15T18:53:11Z

@swift-ci Please test

al45tair · 2026-01-16T14:19:03Z

@swift-ci Please test

al45tair · 2026-01-16T14:54:35Z

@swift-ci Please test Linux platform

al45tair · 2026-01-16T14:54:41Z

@swift-ci Please test macOS platform

al45tair · 2026-01-19T12:27:53Z

@swift-ci Please test

al45tair · 2026-01-19T14:25:51Z

@swift-ci Please test

al45tair · 2026-01-19T14:43:31Z

@swift-ci Please test Linux platform

al45tair · 2026-02-04T12:20:35Z

@swift-ci Please test

al45tair · 2026-02-04T12:21:49Z

I've just checked that this works on ARM64 Windows, since we don't have PR testing coverage there. My small crash test program generated a backtrace :-)

One thing I did spot is that if you install x86-64 Swift on ARM64 Windows, programs seem to crash on exit somewhere in mimalloc. We might need to look into that, but that's nothing to do with this PR.

edymtt

Looks good to me on the build system side (especially the new one) -- not sure if it makes sense to break the rename of SwiftRuntime to SwiftRuntimeCore in the Core build system in a separate PR.

edymtt · 2026-02-04T15:55:17Z

Runtimes/Supplemental/Runtime/CMakeLists.txt

 find_package(SwiftOverlay REQUIRED)
 find_package(SwiftCxxOverlay REQUIRED)
+
+if(APPLE)


The Darwin find module should be already skipping the search if we are not building for Darwin, so this if should not be strictly necessary -- did you hit issues when building at desk?

I think I did, yes, We might want to investigate that, if we're not supposed to need the APPLE condition.

edymtt · 2026-02-04T16:01:09Z

Runtimes/Core/runtime/CMakeLists.txt

+target_compile_definitions(swiftRuntimeCore
  PRIVATE
    -DSWIFT_RUNTIME
    -DSWIFT_TARGET_LIBRARY_NAME=swiftRuntimeCore


Wondering if this is now redundant -- in case, we can remove it at a later time to reduce risks in landing this PR.

I wondered that also, but didn't feel brave enough to delete it :-)

al45tair · 2026-02-04T17:57:36Z

@swift-ci Please test Linux platform

al45tair · 2026-02-05T11:22:00Z

@swift-ci Please test macOS platform

al45tair · 2026-02-05T11:46:55Z

In addition to manually testing crash catching on ARM64 Windows, I've also tested that the Runtime API is able to capture a backtrace with Backtrace.capture(). That is also working.

al45tair · 2026-02-05T16:39:37Z

@swift-ci Please test macOS platform

Don't even bother building the backtracing code if the backtracing code is disabled. rdar://101623384

We were omitting the `end` instruction, the upshot of which was that the compact backtrace was always being decoded as truncated. rdar://101623384

Also, fix two files that managed to use Windows API without importing `WinSDK` as a result of the leakage caused by including `windows.h` from `codeview.h`. rdar://101623384

We should only add the assembly files if we're actually going to build the `Runtime` module. This was tripping up iOS/tvOS/ watchOS builds. rdar://101623384

Tweak things to build on Linux and macOS again. Also fix a few things in the tests. rdar://101623384

We can't use `|| true` on Windows, so use `not` instead. Also, on Windows, program names get downcased because `lit` uses `os.path.normcase()`, so if we have program names in the output, make sure the programs are already lower-case. Fix up the stack overflow test to match the expected output. Plus fix the macOS build to always build for macOS 26.0 or above. rdar://101623384

`not` works on Windows, but not on Linux or Darwin. rdar://101623384

Test for `}` rather than `},`, since we may or may not have extra frames after the ones we expect, depending. rdar://101623384

Tweaked the comment in `Runtime/Config.h`. Fixed a couple of incorrect ARM64 instruction mnemonics. This still needs testing on ARM64 Windows. Fixed an out-of-date comment in `swift-backtrace`. Use a macro in `Backtrace.cpp` to guarantee we don't overrun the buffer, and in the process simplify the code slightly. rdar://101623384

This change in the driver isn't necessary after all. rdar://101623384

I really did mean `ret`. rdar://101623384

We needed to add the `Runtime` module to the new build system, but when I tried doing that, various things broke. Firstly, we ended up with two targets with the name `swiftRuntime`, but only in the static build (because in that build, everything gets pulled in together, so CMake sees both of them). Rename the one that's part of `swiftCore` to `swiftRuntimeCore`. Second, we need some extra paths for the `Cxx` module and its submodules, and since those are part of the `INTERFACE`, we need to make sure we `PUBLIC` link them into the `Cxx` module so that they get passed through when building `swiftRuntime`. rdar://101623384

al45tair · 2026-02-06T08:53:46Z

@swift-ci Please smoke test

…in CI again (#87275) This was fixed in #86694, but the subsequent massive #84906 appears to have [inadvertently disabled adding the needed Android API level](https://github.com/swiftlang/swift/pull/84906/changes#diff-c861caa2fcc08744108e542ca836b07e342f537566e8c5f7a21a0e9125331ed9R2774), as it was started late last year, long before the fix was added.

…in CI again (swiftlang#87275) This was fixed in swiftlang#86694, but the subsequent massive swiftlang#84906 appears to have [inadvertently disabled adding the needed Android API level](https://github.com/swiftlang/swift/pull/84906/changes#diff-c861caa2fcc08744108e542ca836b07e342f537566e8c5f7a21a0e9125331ed9R2774), as it was started late last year, long before the fix was added.

al45tair force-pushed the eng/PR-101623384 branch 2 times, most recently from 4ea287c to d954e8a Compare October 15, 2025 12:25

al45tair force-pushed the eng/PR-101623384 branch from 447ed50 to 435bea0 Compare January 12, 2026 11:07

al45tair force-pushed the eng/PR-101623384 branch 2 times, most recently from d4a345a to 499c6ea Compare January 13, 2026 12:24

al45tair force-pushed the eng/PR-101623384 branch 2 times, most recently from 6512a83 to bb20822 Compare January 14, 2026 11:17

al45tair force-pushed the eng/PR-101623384 branch from 057805c to 51b2102 Compare January 14, 2026 14:16

al45tair force-pushed the eng/PR-101623384 branch from f8e40e5 to 2e707c6 Compare January 15, 2026 18:52

al45tair marked this pull request as ready for review January 20, 2026 12:07

al45tair requested a review from compnerd as a code owner January 20, 2026 12:07

edymtt approved these changes Feb 4, 2026

View reviewed changes

al45tair force-pushed the eng/PR-101623384 branch 2 times, most recently from 6a8a47c to 2ebe6ed Compare February 6, 2026 08:49

al45tair added 12 commits February 6, 2026 08:53

[Backtracing] Don't build backtracing code when it's disabled.

add83b4

Don't even bother building the backtracing code if the backtracing code is disabled. rdar://101623384

[Backtracing] Remember to emit end instruction.

38c4848

We were omitting the `end` instruction, the upshot of which was that the compact backtrace was always being decoded as truncated. rdar://101623384

[Backtracing] Don't pull in windows.h unnecessarily.

f6af22c

Also, fix two files that managed to use Windows API without importing `WinSDK` as a result of the leakage caused by including `windows.h` from `codeview.h`. rdar://101623384

[Backtracing] Fix build failures for places Runtime doesn't build.

58a6470

We should only add the assembly files if we're actually going to build the `Runtime` module. This was tripping up iOS/tvOS/ watchOS builds. rdar://101623384

[Backtracing] Build and test fixes.

d9b0031

Tweak things to build on Linux and macOS again. Also fix a few things in the tests. rdar://101623384

[Backtracing][Test] Use !, not not.

e477882

`not` works on Windows, but not on Linux or Darwin. rdar://101623384

[Backtracing][Tests] We might not have more frames here.

5b4c428

Test for `}` rather than `},`, since we may or may not have extra frames after the ones we expect, depending. rdar://101623384

[Backtracing] We don't need to change the driver.

5438510

This change in the driver isn't necessary after all. rdar://101623384

[Backtracing] Fix silly mistake.

3e7647d

I really did mean `ret`. rdar://101623384

al45tair force-pushed the eng/PR-101623384 branch from 2ebe6ed to 079c165 Compare February 6, 2026 08:53

al45tair merged commit e38c065 into swiftlang:main Feb 6, 2026
3 checks passed

hnrklssn mentioned this pull request Feb 7, 2026

☂️ Use llvm-lit's internal shell by default #84407

Closed

finagolfin mentioned this pull request Feb 9, 2026

[android] Fix CI for recent trunk pulls #87075

Merged

finagolfin mentioned this pull request Feb 17, 2026

[test][android] Fix building the Android tests on the Windows toolchain CI again #87275

Merged

etcwilde mentioned this pull request Feb 20, 2026

utils: enable backtracing in the runtimes build on Windows #87154

Merged

Conversation

al45tair commented Oct 15, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

al45tair commented Oct 15, 2025

Uh oh!

al45tair commented Oct 15, 2025

Uh oh!

al45tair commented Dec 10, 2025

Uh oh!

al45tair commented Jan 12, 2026

Uh oh!

al45tair commented Jan 12, 2026

Uh oh!

al45tair commented Jan 13, 2026

Uh oh!

al45tair commented Jan 13, 2026

Uh oh!

al45tair commented Jan 14, 2026

Uh oh!

al45tair commented Jan 14, 2026

Uh oh!

al45tair commented Jan 14, 2026

Uh oh!

al45tair commented Jan 14, 2026

Uh oh!

al45tair commented Jan 14, 2026

Uh oh!

al45tair commented Jan 15, 2026

Uh oh!

al45tair commented Jan 16, 2026

Uh oh!

al45tair commented Jan 16, 2026

Uh oh!

al45tair commented Jan 16, 2026

Uh oh!

al45tair commented Jan 19, 2026

Uh oh!

al45tair commented Jan 19, 2026

Uh oh!

al45tair commented Jan 19, 2026

Uh oh!

al45tair commented Feb 4, 2026

Uh oh!

al45tair commented Feb 4, 2026

Uh oh!

edymtt left a comment

Choose a reason for hiding this comment

Uh oh!

edymtt Feb 4, 2026

Choose a reason for hiding this comment

Uh oh!

al45tair Feb 4, 2026

Choose a reason for hiding this comment

Uh oh!

edymtt Feb 4, 2026

Choose a reason for hiding this comment

Uh oh!

al45tair Feb 4, 2026

Choose a reason for hiding this comment

Uh oh!

al45tair commented Feb 4, 2026

Uh oh!

al45tair commented Feb 5, 2026

Uh oh!

al45tair commented Feb 5, 2026

Uh oh!

al45tair commented Feb 5, 2026

Uh oh!

al45tair commented Feb 6, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

al45tair commented Oct 15, 2025 •

edited

Loading