Fix build on FreeBSD #38335

drexin · 2021-07-09T19:01:46Z

These changes make Swift build on FreeBSD (tested on 13.0-RELEASE-p1). There are some tests that I had to XFAIL, but the majority works. SourceKit tests are very unstable and seem to fail when too many tests are executed in parallel. Running them in isolation with lit.py -v $SWIFT_BUILD_ROOT/tests-xxx --filter=SourceKit -j12 works, but using more threads causes random failures. The more threads are used, the more tests fail. I have not found the reason, yet.

drexin · 2021-07-09T19:02:03Z

@swift-ci test

include/swift/SILOptimizer/Differentiation/DifferentiationInvoker.h

gottesmm

Some comments.

stdlib/public/Platform/CMakeLists.txt

stdlib/public/Platform/libc-freebsd.modulemap.gyb

test/lit.cfg

test/stdlib/POSIX.swift

unittests/runtime/CMakeLists.txt

swift-ci · 2021-07-09T20:27:39Z

Build failed
Swift Test Linux Platform
Git Sha - 17fc1037a285eef09d4f738a23e9142a876797ce

swift-ci · 2021-07-09T20:45:10Z

Build failed
Swift Test OS X Platform
Git Sha - 17fc1037a285eef09d4f738a23e9142a876797ce

drexin · 2021-07-09T22:30:54Z

@swift-ci test

3405691582 · 2021-07-09T22:40:56Z

stdlib/public/SwiftShims/SwiftStdint.h

@@ -24,7 +24,7 @@

 // Clang has been defining __INTxx_TYPE__ macros for a long time.
 // __UINTxx_TYPE__ are defined only since Clang 3.5.
-#if !defined(__APPLE__) && !defined(__linux__) && !defined(__OpenBSD__)
+#if !defined(__APPLE__) && !defined(__linux__) && !defined(__OpenBSD__) && !defined(FreeBSD)


You probably want defined(__FreeBSD__) here.

Good catch, thanks!

drexin · 2021-07-09T22:55:43Z

@swift-ci test

swift-ci · 2021-07-10T00:07:17Z

Build failed
Swift Test Linux Platform
Git Sha - a8b345b2c6abfa6dc1913ff5f5a25fcc6a7bfe37

drexin · 2021-07-10T00:28:49Z

@swift-ci test linux

swift-ci · 2021-07-10T01:41:31Z

Build failed
Swift Test Linux Platform
Git Sha - a8b345b2c6abfa6dc1913ff5f5a25fcc6a7bfe37

drexin · 2021-07-10T04:06:37Z

@swift-ci clean test linux

finagolfin · 2021-07-10T06:06:04Z

lib/ClangImporter/ClangImporter.cpp

@@ -658,7 +658,7 @@ importer::getNormalInvocationArguments(
    }
  }

-  if (searchPathOpts.SDKPath.empty()) {
+  if (searchPathOpts.SDKPath.empty() && !triple.isOSFreeBSD()) {


Why special case this for FreeBSD, not working?

Yes, it excludes all system headers, causing e.g. the Glibc module to not work. I am unsure why this works on Linux, but not FreeBSD.

You could try adding the ClangImporter flags -Xfrontend -dump-clang-diagnostics -Xcc -v after removing this special case to dump what it's seeing and compare with linux, as I did with SR-14796.

finagolfin · 2021-07-10T06:21:27Z

test/Sanitizers/tsan/tsan.swift

  threads.append(t!)
 #else
  threads.append(t)
 #endif
 }

+print("================ 34 2THRAEDS: \(threads)")


Your debugging info that can be taken out?

finagolfin · 2021-07-10T06:26:00Z

tools/swift-dependency-tool/swift-dependency-tool.cpp

@@ -30,7 +30,7 @@ using namespace fine_grained_dependencies;

 // This introduces a redefinition where ever std::is_same_t<size_t, uint64_t>
 // holds
-#if !(defined(__linux__) || defined(_WIN64))
+#if (!(defined(__linux__) || defined(_WIN64) || defined(__FreeBSD__)))


Added unnecessary first parenthesis.

finagolfin

Good to see this. I don't know about the std::pair changes but otherwise checks out.

finagolfin · 2021-07-10T06:34:26Z

stdlib/public/Platform/CMakeLists.txt

-        FLAGS "-DCMAKE_SDK=${sdk}")
-    list(APPEND glibc_modulemap_target_list ${glibc_header_target})
+    # FreeBSD uses a different module map that does not use this header,
+    # so we don't generate it


Have you tried using this "Glibc" header instead on FreeBSD? I have it almost working for Android, #35707, so I think it should work for the BSDs too.

I tried, but ran into a lot of missing functions. I did not have time to figure out what the problem is, yet, but I will take a look. Thanks for the link. I would like to land this first, though.

drexin · 2021-07-12T21:33:02Z

@swift-ci smoke test

fossdd

Swift on FreeBSD is awesome! Thanks!

drexin · 2021-07-13T23:37:32Z

@swift-ci test macos

drexin · 2021-07-13T23:37:38Z

@swift-ci test windows

drexin · 2021-07-14T00:37:00Z

@swift-ci smoke test

drexin · 2021-07-14T07:23:25Z

@swift-ci test windows

drexin · 2021-07-14T15:27:44Z

@swift-ci test windows

drexin · 2021-07-15T18:39:07Z

@swift-ci test windows

drexin · 2021-07-15T22:20:26Z

@swift-ci test windows

drexin · 2021-07-19T19:02:33Z

@swift-ci test windows

intini · 2021-09-30T23:39:54Z

Hi. I would like to register that I really would like very much to have Swift ported and maintained on FreeBSD.

majortom64 · 2021-10-05T13:38:11Z

Curious what is preventing this from happening? It would be great to be able to deploy Server Side Swift on FreeBSD. Is there anything that can be done to support this effort (financially or other wise)?

Super excited and want to thank everyone who has pushed this along.

erde74 · 2022-01-30T22:40:35Z

@drexin hi, i updated the PR, the patch is based on swift release/5.6 branch. i have test this on FreeBSD 13

FreeBSD lenovo.local 13.0-RELEASE-p6 FreeBSD 13.0-RELEASE-p6 #0: Mon Jan 10 06:28:50 UTC 2022 root@amd64-builder.daemonology.net:/usr/obj/usr/src/amd64.amd64/sys/GENERIC amd64

Testing Time: 5295.68s
Unsupported : 1861
Passed : 6102
Expectedly Failed : 32
Failed : 45
Unexpectedly Passed: 1

freebsd.patch.txt
test-fail.txt

tobiastom · 2022-04-21T09:29:51Z

I would like to second @majortom64 question. @drexin: is there anything we can help to get this merged? Swift on FreeBSD would indeed be awesome.

drexin · 2022-04-21T15:52:26Z

Hey, sorry for the silence here. I was pretty busy with other stuff and this was just a side project. I don't think I will be able to spend any time on this in the near future, but if anyone wants to pick this up, please feel free to.

tobiastom · 2022-04-22T09:40:45Z

@drexin Thank you for the update. I completely understand.

Unfortunately this is beyond by abilities, but hopefully someone else might do it.

b-ncMN · 2022-06-20T16:29:57Z

Hey everyone, coming from #59579
I would be interesting into continuing the work on this

erde74 · 2022-06-21T17:30:58Z

Pleased do Am 20. Juni 2022 18:30:11 MESZ schrieb InfRandomness ***@***.***>:

Hey everyone, coming from #59579 I would be interesting into continuing the work on this -- Reply to this email directly or view it on GitHub: #38335 (comment) You are receiving this because you commented. Message ID: ***@***.***>

-- Diese Nachricht wurde von meinem Android-Gerät mit K-9 Mail gesendet.

b-ncMN · 2022-06-24T05:57:40Z

Hi,
I'm coming back here with a little bit of advancement;
I have made a fork of this official repository, cloned it, made a wip-freebsd branch on top of main and cherry-picked / rebased the wip-freebsd branch from @drexin on my own branch, it is available here
I was wondering whether or not to create a new PR for that new branch?
and I was also wondering about this line specifically,
I have changed it to freebsd-uknown, seeing as how linux is linux-gnu, but I'm not entirely sure and ask for a confirmation

Thanks a lot for the answers in advance ! 😄

arrowd · 2022-06-24T14:48:44Z

include/swift/SILOptimizer/Differentiation/DifferentiationInvoker.h

-    std::pair<ApplyInst *, SILDifferentiabilityWitness *>
-        indirectDifferentiation;
+    ///
+    /// Note: This used to be a std::pair, but on FreeBSD std::pair can't be


I believe, this is not FreeBSD specifics, but a libc++ one. Change the comment accordingly.

FreeBSD specifically disables the copy constructor by setting _LIBCPP_DEPRECATED_ABI_DISABLE_PAIR_TRIVIAL_COPY_CTOR to preserve ABI. It is not a general limitation of libc++.

finagolfin · 2022-06-24T15:20:47Z

I was wondering whether or not to create a new PR for that new branch?

No, first follow the build instructions for linux here, translating to FreeBSD as much as possible, and see if you can bootstrap the compiler and run the validation suite, as has been done for OpenBSD. Only once you're able to build the compiler and run the tests on FreeBSD should you consider submitting a pull.

I was also wondering about this line specifically, I have changed it to freebsd-uknown, seeing as how linux is linux-gnu, but I'm not entirely sure and ask for a confirmation

Yes, that is better, the tests are moving that way, as in #58975.

b-ncMN · 2022-06-24T17:44:53Z

I was wondering whether or not to create a new PR for that new branch?

No, first follow the build instructions for linux here, translating to FreeBSD as much as possible, and see if you can bootstrap the compiler and run the validation suite, as has been done for OpenBSD. Only once you're able to build the compiler and run the tests on FreeBSD should you consider submitting a pull.

I was also wondering about this line specifically, I have changed it to freebsd-uknown, seeing as how linux is linux-gnu, but I'm not entirely sure and ask for a confirmation

Yes, that is better, the tests are moving that way, as in #58975.

Is there a process to follow for fixing LLVM build issues?
Can I provide the changes to https://github.com/apple/llvm-project ? or is it better to propose them upstream directly?

finagolfin · 2022-06-25T00:21:09Z

Is there a process to follow for fixing LLVM build issues?

Yes, you submit it to their Phabricator.

Can I provide the changes to https://github.com/apple/llvm-project ? or is it better to propose them upstream directly?

No, you're supposed to submit changes upstream, not to the Apple fork for Swift.

Bring drexin's FreeBSD support patch up to date with main. Resolved conflicts in prior merge commit and made some minor buildfixes. Still requires some other more general fixes outside this pr to have this build properly at HEAD but have verified this builds OK with those fixes TK on FreeBSD 12.3. Note: I don't intend to support this platform as a priority.

Bring drexin's FreeBSD support patch up to date with main. Resolved conflicts in prior merge commit and made some minor buildfixes. Still requires some other more general fixes outside this pr to have this build properly at HEAD but have verified only this builds OK with those fixes possibly TK including swiftlang#59987 on FreeBSD 12.3. Note: I don't intend to support this platform as a priority.

Bring drexin's FreeBSD support patch up to date with main. Resolved conflicts in prior merge commit and made some minor buildfixes. Still requires some other more general fixes outside this pr to have this build properly at HEAD but have verified only this builds OK with those fixes possibly TK including swiftlang#59987 and bootstrapping disabled on FreeBSD 12.3. Note: I don't intend to support this platform as a priority.

gottesmm reviewed Jul 9, 2021

View reviewed changes

include/swift/SILOptimizer/Differentiation/DifferentiationInvoker.h Outdated Show resolved Hide resolved

gottesmm reviewed Jul 9, 2021

View reviewed changes

drexin force-pushed the wip-freebsd branch from 17fc103 to d281343 Compare July 9, 2021 22:30

3405691582 reviewed Jul 9, 2021

View reviewed changes

drexin force-pushed the wip-freebsd branch 2 times, most recently from cefc502 to a8b345b Compare July 9, 2021 22:55

finagolfin reviewed Jul 10, 2021

View reviewed changes

finagolfin approved these changes Jul 10, 2021

View reviewed changes

finagolfin reviewed Jul 10, 2021

View reviewed changes

drexin force-pushed the wip-freebsd branch from a8b345b to a794e16 Compare July 12, 2021 21:32

fossdd approved these changes Jul 13, 2021

View reviewed changes

Fix build on FreeBSD

43dc209

drexin force-pushed the wip-freebsd branch from a794e16 to 43dc209 Compare July 14, 2021 00:36

lin72h mentioned this pull request Jan 31, 2022

Swift build on FreeBSD ravynsoft/ravynos#209

Open

finagolfin mentioned this pull request Jun 20, 2022

FreeBSD CI #59579

Open

arrowd reviewed Jun 24, 2022

View reviewed changes

MaxDesiatov added platform support FreeBSD FreeBSD platform support labels Jun 15, 2023

Fix build on FreeBSD #38335

Are you sure you want to change the base?

Fix build on FreeBSD #38335

Conversation

drexin commented Jul 9, 2021

drexin commented Jul 9, 2021

gottesmm left a comment

Choose a reason for hiding this comment

swift-ci commented Jul 9, 2021

swift-ci commented Jul 9, 2021

drexin commented Jul 9, 2021

Choose a reason for hiding this comment

Choose a reason for hiding this comment

drexin commented Jul 9, 2021

swift-ci commented Jul 10, 2021

drexin commented Jul 10, 2021

swift-ci commented Jul 10, 2021

drexin commented Jul 10, 2021

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

finagolfin Jul 10, 2021 • edited Loading

Choose a reason for hiding this comment

finagolfin left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

drexin commented Jul 12, 2021

fossdd left a comment

Choose a reason for hiding this comment

drexin commented Jul 13, 2021

drexin commented Jul 13, 2021

drexin commented Jul 14, 2021

drexin commented Jul 14, 2021

drexin commented Jul 14, 2021

drexin commented Jul 15, 2021

drexin commented Jul 15, 2021

drexin commented Jul 19, 2021

intini commented Sep 30, 2021

majortom64 commented Oct 5, 2021

erde74 commented Jan 30, 2022

tobiastom commented Apr 21, 2022

drexin commented Apr 21, 2022

tobiastom commented Apr 22, 2022

b-ncMN commented Jun 20, 2022 • edited Loading

erde74 commented Jun 21, 2022 via email

b-ncMN commented Jun 24, 2022

Choose a reason for hiding this comment

Choose a reason for hiding this comment

finagolfin commented Jun 24, 2022

b-ncMN commented Jun 24, 2022

finagolfin commented Jun 25, 2022

finagolfin Jul 10, 2021 •

edited

Loading

b-ncMN commented Jun 20, 2022 •

edited

Loading