I also ran the test suite for the git master (but did not install). tramp-test30-make-auto-save-file-name is failing. The stdout+err of the test run is also attached. Let me know if you want to see more verbose test output and how to enable that.
> Yes, I killed the process because it hangs indefinitely there. The last
> minibuffer msg is:
> "Tramp: Decoding local file '/tmp....' with 'base64-decode-region'
> Yes the chown operation is successful on the local cygwin host. (But not on
> the remote, if that matters.)
... but it hangs when you are using the plink method. I have wrapped it
with a timeout of 5 seconds; could you please check whether it works
better now for you? (You must pull from the repo, again.)
> The 2.2.12-pre tests now look good on the latest Cygwin emacs.
> Ran 42 tests, 42 results as expected (2015-05-31 19:10:21-0400)
> 2 expected failures
Re: tramp hangs decoding local file (cygwin, emacs-w32)
Still hangs at "decoding local file" even with the timeout fix. But this is looking more like a cygwin problem.
My baseline, working setup is:
emacs 24.4.1, tramp 2.2.9-24.4
Changing only emacs and tramp to the latest available in the cygwin repo:
emacs 24.5.1, tramp 2.2.11-24.5 ... plink tramp sessions continue to work ok. And using your latest 2.2.12-pre with either setup works fine.
The trouble starts when the Cygwin system is updated. In my case, the cygwin engine goes from 1.7.35-1 to 2.0.2-1 .. and a bunch of other stuff comes with that which I haven't taken the time to A/B test. The point is, that something in that update that is not emacs/tramp/plink, corresponds to the trouble.
Your 2.2.12-pre expected tests are passing on all my Cygwin setups. But the timeout code does not unstick the hanging process on cygwin 2.0.2-1.
Thanks for looking at this from the tramp side. Sorry I wasn't more thorough isolating the problem.
> My baseline, working setup is:
> cygwin 1.7.35-1
> emacs 24.4.1, tramp 2.2.9-24.4
> plink 0.64
Unfortunately, I cannot test this config. I don't use MS Windows machines.
> Your 2.2.12-pre expected tests are passing on all my Cygwin setups. But the
> timeout code does not unstick the hanging process on cygwin 2.0.2-1.
A very recent patch to Tramp 2.2.12-pre adds more traces in case you
cancel Tramp with C-g. Maybe you could pull it, and run your test with
tramp-verbose 10. There is a minor chance we get more information.
> Thanks for looking at this from the tramp side. Sorry I wasn't more thorough
> isolating the problem.
No problem, it's my daily fun to understand what Tramp users are doing :-)
> I'll report the problem to the cygwin mail list sometime this week. Let me
> know if the debug output provides new insights.
Unfortunately not. I made a small misthinking when adding the traces:
they report the status on the remote host when you interrupt Tramp; but
in your case we have a blocking call on the local host, sending the
Well, I put this on hold. Let me know, if the cygwin maintainers need
more help for analyzing.