mercurial commit fails with large files

[Andrew]

Binary package hint: mercurial

When attempting the following commands:
hg init
hg ci -Aym "Initial"

and one of the files being added is over 10MB, mercurial errors due to a python code bug:

somelargfile.tar.gz: files over 10MB may cause memory and performance problems
(use 'hg revert somelargfile.tar.gz' to unadd the file)
Traceback (most recent call last):
  File "/usr/bin/hg", line 20, in <module>
    mercurial.dispatch.run()
  File "/var/lib/python-support/python2.5/mercurial/dispatch.py", line 20, in run
    sys.exit(dispatch(sys.argv[1:]))
  File "/var/lib/python-support/python2.5/mercurial/dispatch.py", line 29, in dispatch
    return _runcatch(u, args)
  File "/var/lib/python-support/python2.5/mercurial/dispatch.py", line 93, in _runcatch
    ui.warn(_("abort: %s\n") % inst[1])
IndexError: tuple index out of range

Ubuntu: Intrepid
Architecture: amd64
Mercurial Version: 1.0.1-5.1

[Andrew (andrew-rw-robinson)]

Even running "hg st" after this happens failes with the same error

[Andrew (andrew-rw-robinson)]

Changing:
    except socket.error, inst:
        ui.warn(_("abort: %s\n") % inst[1])
to:
    except socket.error, inst:
        ui.warn(_("abort: %s\n") % inst.message)

I then get:
abort: AF_UNIX path too long

Which is much more helpful of an error, So I guess hg has a problem with long path names

[Andrew (andrew-rw-robinson)]

Upstream bug:
http://www.selenic.com/mercurial/bts/issue1208

[Andrew (andrew-rw-robinson)]

Work-around noted in that bug:

when I comment out the "hgext.inotify =" line of /etc/mercurial/hgrc.d/hgext.rc,
that makes the test pass. So it *is* inotify-specific.

[tonfa (bboissin)]

It is really bad if the inotify extension is enabled globally. Did you do that or was it enabled by ubuntu ?

[Andrew (andrew-rw-robinson)]

It was enabled by default (globally) on ubuntu install.

I would suggest as a temporary "fix" for this bug that ubuntu should ship it disabled

[tonfa (bboissin)]

On Fri, Nov 14, 2008 at 06:26:21PM -0000, Andrew wrote:
> It was enabled by default (globally) on ubuntu install.
>
> I would suggest as a temporary "fix" for this bug that ubuntu should
> ship it disabled

All extensions should be disabled by default, that's what the recent
debian package is doing (I thought it propagated to ubuntu).

--
:wq

[Vernon Tang (vtang)]

 affects ubuntu/mercurial
 status fixreleased

[Andrew (andrew-rw-robinson)]

Thank you for fixing this. Sorry to be ignorant, but how do I find the version the fix would be found in?