Signed-off-by: Eric S. Raymond <esr@thyrsus.com>

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

---
for non-ascii input, irkerhook.py dies with:

remote: Traceback (most recent call last):
remote:   File "/usr/bin/irkerhook.py", line 484, in <module>
remote:     ship(extractor, commit, not notify)
remote:   File "/usr/bin/irkerhook.py", line 391, in ship
remote:     metadata = extractor.commit_factory(commit)
remote:   File "/usr/bin/irkerhook.py", line 262, in commit_factory
remote:     metainfo = do("git log -1 '--pretty=format:%an <%ae>|%s' " +
shellquote(commit.commit))
remote:   File "/usr/bin/irkerhook.py", line 50, in do
remote:     return unicode(commands.getstatusoutput(command)[1],
locale.getpreferredencoding() or 'UTF-8')
remote: UnicodeDecodeError: 'ascii' codec can't decode byte 0xc3 in
position 8: ordinal not in range(128)

for e.g. this author name:
$ git log -n1 --pretty="format:%an" | cat;echo
Timo Teräs
$ git log -n1 --pretty="format:%an" | od -x;echo
0000000 6954 6f6d 5420 7265 a4c3 0073
0000013
So:
$ for i in C POSIX en_US.UTF-8;do echo "# $i";LC_ALL=$i python -c
"import locale;print locale.getlocale()[1];print unicode(\"ä\",
locale.getlocale()[1] or 'UTF-8').encode(locale.getlocale()[1] or
'UTF-8')";done
None
ä
None
ä
None
ä

The constructed mercurial url is wrong - it creates urls like
"repo/rev3456fg", which just get redirected to the default log, rather
than the commit. The url should be "repo/rev/3456fg".

Signed-off-by: Eric S. Raymond <esr@thyrsus.com>

This does not change anything for tinyurl (the default), but allows
using http://tinyarro.ws/api-create.php?utfpure=1&url=


In case invalid characters are encountered, fall back to str.

Signed-off-by: Eric S. Raymond <esr@thyrsus.com>

Signed-off-by: Eric S. Raymond <esr@thyrsus.com>

AI0867 reported this was causing a rare hang during disconnect.
Also, set the connection status to expired if the thread terminates
due to exception.

Boom!  Blew up, first time tested.

Alas, the fix violates RFC2812.

Usage: outcmd|irk target -
Sending "-" is still possible by echo -|irk target -

Signed-off-by: Eric S. Raymond <esr@thyrsus.com>

Signed-off-by: Eric S. Raymond <esr@thyrsus.com>

This fixes a corner case where if you had a non-ASCII character in a
commit and a filter command, you would end up with an unicode string and
irkerhook.py would crash.
But it would somehow work if you did not have a filter command.
For both modes cases to work, forcing unicode in do() and then using
unicode methods are required.
JSON is unicode by default, and this is why the issues happened.

Signed-off-by: Eric S. Raymond <esr@thyrsus.com>