linux: rewrite mime parser

linux: rewrite mime parser

This is based on the logic in the Firefox code.
It passes existing tests and one new test.

TEST=covered by new unit test

[Evan Martin, 2009-09-09 00:03:16]

hold on a sec, this is still wrong

[Evan Martin, 2009-09-09 00:11:52]

ok, please review now

[piman, 2009-09-09 00:17:00]

http://codereview.chromium.org/200052/diff/1002/9
File webkit/glue/plugins/plugin_lib_linux.cc (right):

http://codereview.chromium.org/200052/diff/1002/9#newcode115
Line 115: bool PluginLib::ParseMIMEDescription(
This function doesn't fail any more. change it to return void ?

http://codereview.chromium.org/200052/diff/1002/9#newcode131
Line 131: end = description.find(':', ofs);
suggestion: declare 'end' here instead of above. it makes it clear that end is
not re-used across the loop iterations.

http://codereview.chromium.org/200052/diff/1002/9#newcode135
Line 135: ofs = end + 1;
what if ofs == description.length() now ? (e.g. you're parsing "mime/type:")
I'm not sure that string::find handles that gracefully. Worth adding a test for
this case.

http://codereview.chromium.org/200052/diff/1002/9#newcode142
Line 142: ofs = end + 1;
same here, what if you're parsing "mime/type:ext,ex2:" ?

http://codereview.chromium.org/200052/diff/1002/10
File webkit/glue/plugins/plugin_lib_unittest.cc (right):

http://codereview.chromium.org/200052/diff/1002/10#newcode79
Line 79: ASSERT_EQ(7U, types.size());
Worth a spot check that the MIME types with a ';' in them are coming out fine.

[Evan Martin, 2009-09-09 00:30:18]

Ok, I think I've addressed all your comments (mostly adding more tests).

[piman, 2009-09-09 01:09:53]

LGTM