| Message ID | pull.1074.git.1636542550889.gitgitgadget@gmail.com (mailing list archive) |
|---|---|
| State | Accepted |
| Commit | 974ef7ced24a782ff32b2248103e25a397276f36 |
| Headers | show |
| Series | simple-ipc: work around issues with Cygwin's Unix socket emulation | expand |
"Johannes Schindelin via GitGitGadget" <gitgitgadget@gmail.com> writes: > From: Johannes Schindelin <johannes.schindelin@gmx.de> > > Cygwin emulates Unix sockets by writing files with custom contents and > then marking them as system files. > > The tricky problem is that while the file is written and its `system` > bit is set, it is still identified as a file. This caused test failures > when Git is too fast looking for the Unix sockets and then complains > that there is a plain file in the way. > > Let's work around this by adding a delayed retry loop, specifically for > Cygwin. > > Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de> > --- OK, I was about to ask for an Ack from Cygwin folks but I see the original that got Ack is more-or-less the same except for placement of the comments (and this version of course is more polished than the "this should work---please try it" version), so let me pretend that this got tested-by from those who were happy with the original in the old thread. Thanks. > simple-ipc: work around issues with Cygwin's Unix socket emulation > > Adam Dinwoodie reported problems running the simple-ipc tests on Cygwin > [https://lore.kernel.org/git/20211104194619.GA12886@dinwoodie.org]. This > patch works around the underlying problem, which is rooted in Cygwin's > implementation details. > > With this patch, I could not reproduce the problem, even with sh > t0052-simple-ipc.sh --stress-limit=30. > > As per Junio's encouragement > [https://lore.kernel.org/git/xmqqee7ozyx4.fsf@gitster.g], I am > submitting this still in the -rc phase, hoping that it will make it into > v2.34.0 final. > > Published-As: https://github.com/gitgitgadget/git/releases/tag/pr-1074%2Fdscho%2Fcygwin-vs-simple-ipc-workaround-v1 > Fetch-It-Via: git fetch https://github.com/gitgitgadget/git pr-1074/dscho/cygwin-vs-simple-ipc-workaround-v1 > Pull-Request: https://github.com/gitgitgadget/git/pull/1074 > > compat/simple-ipc/ipc-unix-socket.c | 22 ++++++++++++++++++++++ > 1 file changed, 22 insertions(+) > > diff --git a/compat/simple-ipc/ipc-unix-socket.c b/compat/simple-ipc/ipc-unix-socket.c > index 4e28857a0a1..28a79289d4f 100644 > --- a/compat/simple-ipc/ipc-unix-socket.c > +++ b/compat/simple-ipc/ipc-unix-socket.c > @@ -35,6 +35,28 @@ enum ipc_active_state ipc_get_active_state(const char *path) > } > } > > +#ifdef __CYGWIN__ > + /* > + * Cygwin emulates Unix sockets by writing special-crafted files whose > + * `system` bit is set. > + * > + * If we are too fast, Cygwin might still be in the process of marking > + * the underlying file as a system file. Until then, we will not see a > + * Unix socket here, but a plain file instead. Just in case that this > + * is happening, wait a little and try again. > + */ > + { > + static const int delay[] = { 1, 10, 20, 40, -1 }; > + int i; > + > + for (i = 0; S_ISREG(st.st_mode) && delay[i] > 0; i++) { > + sleep_millisec(delay[i]); > + if (lstat(path, &st) == -1) > + return IPC_STATE__INVALID_PATH; > + } > + } > +#endif > + > /* also complain if a plain file is in the way */ > if ((st.st_mode & S_IFMT) != S_IFSOCK) > return IPC_STATE__INVALID_PATH; > > base-commit: 6c220937e2b26d85920bf2d38ff2464a0d57fd6b
On Wed, 10 Nov 2021 at 17:11, Junio C Hamano <gitster@pobox.com> wrote: > > "Johannes Schindelin via GitGitGadget" <gitgitgadget@gmail.com> > writes: > > > From: Johannes Schindelin <johannes.schindelin@gmx.de> > > > > Cygwin emulates Unix sockets by writing files with custom contents and > > then marking them as system files. > > > > The tricky problem is that while the file is written and its `system` > > bit is set, it is still identified as a file. This caused test failures > > when Git is too fast looking for the Unix sockets and then complains > > that there is a plain file in the way. > > > > Let's work around this by adding a delayed retry loop, specifically for > > Cygwin. > > > > Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de> > > --- > > OK, I was about to ask for an Ack from Cygwin folks but I see the > original that got Ack is more-or-less the same except for placement > of the comments (and this version of course is more polished than > the "this should work---please try it" version), so let me pretend > that this got tested-by from those who were happy with the original > in the old thread. Late to the party – other bits of life got in the way – but I've just confirmed this appears to resolve the issue for me too. Adam
diff --git a/compat/simple-ipc/ipc-unix-socket.c b/compat/simple-ipc/ipc-unix-socket.c index 4e28857a0a1..28a79289d4f 100644 --- a/compat/simple-ipc/ipc-unix-socket.c +++ b/compat/simple-ipc/ipc-unix-socket.c @@ -35,6 +35,28 @@ enum ipc_active_state ipc_get_active_state(const char *path) } } +#ifdef __CYGWIN__ + /* + * Cygwin emulates Unix sockets by writing special-crafted files whose + * `system` bit is set. + * + * If we are too fast, Cygwin might still be in the process of marking + * the underlying file as a system file. Until then, we will not see a + * Unix socket here, but a plain file instead. Just in case that this + * is happening, wait a little and try again. + */ + { + static const int delay[] = { 1, 10, 20, 40, -1 }; + int i; + + for (i = 0; S_ISREG(st.st_mode) && delay[i] > 0; i++) { + sleep_millisec(delay[i]); + if (lstat(path, &st) == -1) + return IPC_STATE__INVALID_PATH; + } + } +#endif + /* also complain if a plain file is in the way */ if ((st.st_mode & S_IFMT) != S_IFSOCK) return IPC_STATE__INVALID_PATH;