When I tried to rebase with the following command: hg rebase -k -b . -d 138f65958545 --debug I got the follwing error: dom/ipc/TabParent.cpp: versions differ -> m (premerge) Files to be: dom/ipc/TabParent.cpp dom/ipc/TabParent.cpp dom/ipc/TabParent.cpp running e:\m\mozilla-unified\mach clang-format --assume-filename $HG_FILENAME -p c:\users\kimu\appdata\local\temp\tmpet53ir.cpp abort: clang-format: mach exited with status 1 Do I have to manually resolve the conflict? It is really painful.

Thanks for running this id debug mode! Please copy 'dom/ipc/TabParent.cpp' from the base revision to the destination revision, overwrite it and run: >>mach clang-format --assume-filename dom/ipc/TabParent.cpp -p dom/ipc/TabParent.cpp Please copy and paste here the output. Also do did you run 'mach bootstrap', do you have in ~/.mozbuild/clang-tools/ ?

(In reply to Andi-Bogdan Postelnicu [:andi] from comment #1) > Please copy 'dom/ipc/TabParent.cpp' from the base revision to the > destination revision, overwrite it and run: > > >>mach clang-format --assume-filename dom/ipc/TabParent.cpp -p dom/ipc/TabParent.cpp > > Please copy and paste here the output. I attached the output because it was large. > Also do did you run 'mach bootstrap', do you have in > ~/.mozbuild/clang-tools/ ? Yes, I re-ran `./mach bootstrap` just in case. I also confirmed that I have `~/.mozbuild/clang-tools/` folder. By the way, my `.mozbuild` is an NTFS junction to another drive because my system drive is a small capacity SSD. Is it relevant?

Can you upload a remote bookmark with the changeset that needs to be rebased?

https://hg.mozilla.org/try/rev/2d4d022a4a7251bf6598f23ba2f0619676d0f656 If this is not what you want, please teach me a command.

Are you sure that’s the revision since I don’t see the problematic cpp inside it?

I modified format-source to see error messages from mach: --- a/hgext/format-source/__init__.py +++ b/hgext/format-source/__init__.py @@ -235,17 +235,21 @@ def run_tools(ui, root, tool, cmd, filep ui.debug('running %s\n' % format_cmd) ui.pushbuffer(subproc=True) try: ui.system(format_cmd, environ=env, cwd=root, onerr=error.Abort, errprefix=tool) - finally: + except: + newcontent = ui.popbuffer() + ui.debug('result %s\n' % newcontent) + raise + else: newcontent = ui.popbuffer() return newcontent def touched(repo, old_ctx, new_ctx, paths): matcher = rootedmatch(repo, new_ctx, paths) if any(path in new_ctx for path in paths): status = old_ctx.status(other=new_ctx, match=matcher) return bool(status.modified or status.added) Result: dom/ipc/TabParent.cpp: versions differ -> m (premerge) Files to be: dom/ipc/TabParent.cpp dom/ipc/TabParent.cpp dom/ipc/TabParent.cpp running e:\m\mozilla-unified\mach clang-format --assume-filename $HG_FILENAME -p c:\users\kimu\appdata\local\temp\tmpyuyb9r.cpp result 'e:\m\mozilla-unified\mach' は、内部コマンドまたは外部コマンド、 操作可能なプログラムまたはバッチ ファイルとして認識されていません。 (English: 'e:\m\mozilla-unified\mach' is not recognized as an internal or external command, operable program or batch file.) I added following settings to ~/.hgrc: [format-source] clang-format = /e/m/mozilla-unified/mach clang-format --assume-filename $HG_FILENAME -p clang-format:configpaths = .clang-format, .clang-format-ignore clang-format:fileext = .cpp, .c, .h Then I got the following error: dom/ipc/TabParent.cpp: versions differ -> m (premerge) Files to be: dom/ipc/TabParent.cpp dom/ipc/TabParent.cpp dom/ipc/TabParent.cpp running /e/m/mozilla-unified/mach clang-format --assume-filename $HG_FILENAME -p c:\users\kimu\appdata\local\temp\tmpeaopjl.cpp result 指定されたパスが見つかりません。 (English: The system cannot find the file specified.) Maybe this bug is Windows-specific and Windows-to-UNIX (mingw) path conversions are required?

On Windows, the command processor (cmd.exe) has no way to parse and execute shbang, so it could not run mach directly. This setting worked for me: [format-source] clang-format = sh -c mach clang-format --assume-filename $HG_FILENAME -p clang-format:configpaths = .clang-format, .clang-format-ignore clang-format:fileext = .cpp, .c, .h But I found another (probably plaform-independent) problem. I'll file another bug.

(In reply to Masatoshi Kimura [:emk] from comment #7) > But I found another (probably plaform-independent) problem. I'll file > another bug. Filed bug 1511668.

(In reply to undefined from comment #undefined) > Don’t you run from the build environment? Are you running from cmd directly?

(In reply to Andi-Bogdan Postelnicu [:andi] from comment #9) > (In reply to undefined from comment #undefined) > > > Don’t you run from the build environment? Are you running from cmd directly? I run `hg rebase` from the msys shell. But apparently ui.system() invokes the command via cmd on Windows. I have no control over that.

This is very strange, is your setup custom? I'm asking this because other engineers have used this on windows without any issues.

As I already said, I installed files to D: as much as possible. My objdir is on an exFAT drive. I did not enalbe fsmonitor because it only times out on my environment.

I have the same problem and I'm using the normal MozillaBuild 3.2 build shell (start-shell.bat).

Did the fix from :emk work?

I no longer get the clang-format error, but instead some of my changesets were dropped (which I guess could be bug 1511668).

Actually even worse - in my case it dropped *all* changes to .cpp/.h files.

Can you poște onto the new bug a log? Does it have the same behavior on other platforms?

(In reply to Andi-Bogdan Postelnicu [:andi] from comment #11) > I'm asking this because other engineers have used this on windows without any issues. Are they using native win32 mercurial? I heard Windows developers had to use WSL to deal with Phacricator problems.

You might have a point with the WSL, thanks for mentioning this. I will come up to you with a resolution, will try to reproduce this in a win vm.

(In reply to Andi-Bogdan Postelnicu [:andi] from comment #17) > Can you poște onto the new bug a log? Does it have the same behavior on > other platforms? Posted my patch set to bug 1511668 as well. (In reply to Masatoshi Kimura [:emk] from comment #18) > Are they using native win32 mercurial? I heard Windows developers had to use > WSL to deal with Phacricator problems. Good point, I'm not using WSL, either (and couldn't even if I wanted to, because my desktop is still on Windows 7).

Pushed by cosheehan@mozilla.com: https://hg.mozilla.org/hgcustom/version-control-tools/rev/42179015d473 extensions: Improve hg-format source with default parameters on windows r=sheehan

I got `Command Not Found: machclang-format` error with this patch. Proposed fix: --- a/hgext/format-source/__init__.py +++ b/hgext/format-source/__init__.py @@ -90,17 +90,17 @@ if util.safehasattr(registrar, 'configit file_storage_path = '.hg-format-source' def return_default_clang_format(repo): arguments = ['clang-format', '--assume-filename', '$HG_FILENAME', '-p'] # On windows we need this to call the command in a shell, see Bug 1511594 if os.name == 'nt': - clang_format_cmd = 'sh mach' + ' '.join(arguments) + clang_format_cmd = 'sh mach ' + ' '.join(arguments) else: clang_format_cmd = os.path.join(repo.root, "mach") + ' '.join(arguments) clang_format_cfgpaths = ['.clang-format', '.clang-format-ignore'] clang_fortmat_fileext = ('.cpp', '.c', '.cc', '.h') return clang_format_cmd, clang_format_cfgpaths, clang_fortmat_fileext

Right good catch, otherwise did it work with your fix?

Yes, rebase succeeded (including all .cpp files) with my fix. Thanks for the great work!

Pushed by cosheehan@mozilla.com: https://hg.mozilla.org/hgcustom/version-control-tools/rev/efdd5165c82e extensions: hg-format add missing space between command and arguments r=sheehan

(In reply to Masatoshi Kimura [:emk] from comment #26) > Yes, rebase succeeded (including all .cpp files) with my fix. Thanks for the > great work! Seconded, thanks as well.