| Submitter | Yuya Nishihara |
|---|---|
| Date | May 7, 2018, 1:11 p.m. |
| Message ID | <ea801aa5d559e37f68fb.1525698683@mimosa> |
| Download | mbox | patch |
| Permalink | /patch/31300/ |
| State | Accepted |
| Headers | show |
Comments
On Mon, 07 May 2018 09:11:23 -0400, Yuya Nishihara <yuya@tcha.org> wrote: > # HG changeset patch > # User Yuya Nishihara <yuya@tcha.org> > # Date 1521963324 -32400 > # Sun Mar 25 16:35:24 2018 +0900 > # Node ID ea801aa5d559e37f68fb9bdda47dec6e58abcb88 > # Parent 2be95f3cc4f8a4b59f7c50af093ea033d1d09c5e > sshserver: do setbinary() by caller (API) This seems to be the cause of the recent odd test failures on Windows. I'm not in front of the testbot machine, but test-http-branchmap.t is hanging on my laptop here: SALT1526160986 87 0 + hg -R b ci -m baz + echo SALT1526160986 88 0 SALT1526160986 88 0 + hg push -R b -e '"c:/Python27/python.exe" oldhg' ssh://dummy/ --encoding latin1 For whatever reason, -t30 doesn't cause the test to timeout. If I kill the python.exe doing the push, the test passes(!), and the server shuts down properly. I suspect a flush is missing, but I've no idea where.
On Sat, 12 May 2018 17:49:30 -0400, Matt Harbison wrote: > On Mon, 07 May 2018 09:11:23 -0400, Yuya Nishihara <yuya@tcha.org> wrote: > > > # HG changeset patch > > # User Yuya Nishihara <yuya@tcha.org> > > # Date 1521963324 -32400 > > # Sun Mar 25 16:35:24 2018 +0900 > > # Node ID ea801aa5d559e37f68fb9bdda47dec6e58abcb88 > > # Parent 2be95f3cc4f8a4b59f7c50af093ea033d1d09c5e > > sshserver: do setbinary() by caller (API) > > This seems to be the cause of the recent odd test failures on Windows. > I'm not in front of the testbot machine, but test-http-branchmap.t is > hanging on my laptop here: > > SALT1526160986 87 0 > + hg -R b ci -m baz > + echo SALT1526160986 88 0 > SALT1526160986 88 0 > + hg push -R b -e '"c:/Python27/python.exe" oldhg' ssh://dummy/ --encoding > latin1 Fixed. I'll send the patches shortly. > For whatever reason, -t30 doesn't cause the test to timeout. If I kill > the python.exe doing the push, the test passes(!), and the server shuts > down properly. I suspect a flush is missing, but I've no idea where. Some blocking I/O can't be interrupted on Windows.
Patch
diff --git a/contrib/hg-ssh b/contrib/hg-ssh --- a/contrib/hg-ssh +++ b/contrib/hg-ssh @@ -43,6 +43,9 @@ from mercurial import ( ) def main(): + # Prevent insertion/deletion of CRs + dispatch.initstdio() + cwd = os.getcwd() readonly = False args = sys.argv[1:] diff --git a/mercurial/dispatch.py b/mercurial/dispatch.py --- a/mercurial/dispatch.py +++ b/mercurial/dispatch.py @@ -83,7 +83,7 @@ class request(object): def run(): "run the command in sys.argv" - _initstdio() + initstdio() req = request(pycompat.sysargv[1:]) err = None try: @@ -112,7 +112,7 @@ def run(): sys.exit(status & 255) if pycompat.ispy3: - def _initstdio(): + def initstdio(): pass def _silencestdio(): @@ -132,7 +132,7 @@ if pycompat.ispy3: except IOError: pass else: - def _initstdio(): + def initstdio(): for fp in (sys.stdin, sys.stdout, sys.stderr): procutil.setbinary(fp) diff --git a/mercurial/wireprotoserver.py b/mercurial/wireprotoserver.py --- a/mercurial/wireprotoserver.py +++ b/mercurial/wireprotoserver.py @@ -798,10 +798,6 @@ class sshserver(object): hook.redirect(True) ui.fout = repo.ui.fout = ui.ferr - # Prevent insertion/deletion of CRs - procutil.setbinary(self._fin) - procutil.setbinary(self._fout) - def serve_forever(self): self.serveuntil(threading.Event()) sys.exit(0)