[darcs-users] darcs failed: (sftp) failed to fetch files.
David Roundy
droundy at darcs.net
Wed Sep 7 12:06:25 UTC 2005
On Tue, Sep 06, 2005 at 05:26:02PM -0500, Michael Shuler wrote:
> Keyed password-less get does work successfully:
Great! I think this is probably why there haven't been earlier reports of
this problem, since it's generally best to use darcs with a password-less
ssh setup.
I'm wondering if perhaps the bug is similar to the problem we've had with
calling nvi, and perhaps the same fix will work. Can you try the attached
patch? This sets the stdin file descriptor to "non-blocking read" before
calling external programs. I have no idea if this'll work, but I also
don't have any other ideas for a solution.
If this *doesn't* work, we may have to leave this for now, since I really
don't know how to debug problems like this (which is why the nvi bug--which
I'm hoping this is similar to--had to wait until Benedikt figured it out).
In which case it would be great if you filed a bug report with the BTS.
> darcs failed: (sftp) failed to fetch files.
> source directory: /var/lib/darcs/rs-rhn-api/_darcs/patches
> source files:
> 20050824212913-0090d-f8af9b6acc8d3da0868bc2b391284072d53dcfee.gz
> sftp output:
> Permission denied (publickey,password,keyboard-interactive).
> Couldn't read packet: Connection reset by peer
This is a bit odd. Why would the connection be reset by peer? Could it be
that ssh is timing out waiting for the password?
Also, on the workaround side, I'm not sure if you're familiar with
ssh-agent, but if you use one, password-less public key logins needn't be
insecure.
--
David Roundy
http://www.darcs.net
-------------- next part --------------
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
New patches:
[possible fix for ssh password-reading troubles on sid.
David Roundy <droundy at darcs.net>**20050907115400]
<
> {
hunk ./Exec.lhs 53
careful_unwords (a:as) = "\""++a++"\" "++ careful_unwords as
careful_unwords [] = ""
#else
- -exec c args inp out = do
+exec c args inp out = with_noblocking_stdin $ do
fval <- c_fork
case fval of
-1 -> return $ ExitFailure $ 1
hunk ./Exec.lhs 87
#endif
exec_interactive :: String -> [String] -> IO ExitCode
+exec_interactive arg0 args = with_noblocking_stdin $ rawSystem arg0 args
hunk ./Exec.lhs 89
+with_noblocking_stdin :: IO a -> IO a
#ifndef WIN32
hunk ./Exec.lhs 91
- -
- -exec_interactive arg0 args = do
+with_noblocking_stdin j = do
stdin `seq` return ()
bracket (setFdOption stdInput NonBlockingRead False)
(\_ -> setFdOption stdInput NonBlockingRead True)
hunk ./Exec.lhs 95
- - (\_ -> rawSystem arg0 args)
- -
+ (\_ -> j)
#else
hunk ./Exec.lhs 97
- -
- -exec_interactive = rawSystem
- -
+with_noblocking_stdin = id
#endif
\end{code}
}
Context:
[add changelog entry for configure script checking on darcs being present.
David Roundy <droundy at darcs.net>**20050905113258]
[fix bug where we tried to run darcs when darcs wasn't present.
David Roundy <droundy at darcs.net>**20050905112935]
[revert accidental directory name change in Test.
David Roundy <droundy at darcs.net>**20050904123424]
[fix bug triggered by --no-pristine-tree and running test.
David Roundy <droundy at darcs.net>**20050903132055
The trouble was that the NoPristine version of createPristineDirectoryTree
would fail if the directory already exists, which isn't the intended
behavior. I also took this opportunity to remove the "stubbornly" function
and replace some stubborn directory creation with
createDirectoryIfMissing.
]
[don't create test directory if we don't want to actually run test.
David Roundy <droundy at darcs.net>**20050903130722]
[add test script that displays --no-pristine test-related bug.
David Roundy <droundy at darcs.net>**20050903132906]
[add changelog entry for recent pristine bugfix.
David Roundy <droundy at darcs.net>**20050903134039]
[clean up docs on flags directly to darcs (not to a darcs command).
David Roundy <droundy at darcs.net>**20050903124050]
[bump version to 1.0.4rc1.
David Roundy <droundy at darcs.net>**20050903114002]
[update the web page to direct new users first to the precompiled binaries rather than first to the source
zooko at zooko.com**20050902162737]
[TAG 1.0.4pre4
David Roundy <droundy at darcs.net>**20050901110418]
Patch bundle hash:
c36c8ff4c27bbce3759bcf3bf453847595f95fe8
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)
iD8DBQFDHtWtQ6uZI9PVvOwRAlWJAKD9+g83rKlPerrYaGS/jpK8WvJ6RwCdGmzP
UTvf/PNa5zoRnzKjbBQ5k2Q=
=mrGT
-----END PGP SIGNATURE-----
More information about the darcs-users
mailing list