Leif Hedstrom created TS-4926:
---------------------------------
Summary: Optimize and improve mime_scanner_get() and
mime_parser_parse()
Key: TS-4926
URL: https://issues.apache.org/jira/browse/TS-4926
Project: Traffic Server
Issue Type: Improvement
Components: HTTP
Reporter: Leif Hedstrom
It seems there's some improvements to be done in the interaction between e.g.
mime_scanner_get() and mime_parser_parse(). For example:
1) mime_scanner_get() does not recognize the ':' as it's walking the string
looking for cr/lf's, WS's for continuations etc. Seems like it'd make sense
here to also mark the position of the ':', such that the caller doesn't have to
look for that again (mime_parser_parse() will call memchr() to find it).
2) mime_scanner_get() doesn't deal with various illegal characters, either
leaving it up to the caller(s) to deal with, or not being dealt with at all. It
seems it'd be pretty easy to include the checks for that here, and then avoid
any checks further up in the calling chain entirely.
3) mime_scanner_parse() does not properly deal with white spaces before the
':', which is a protocol error (instead we just ignore 'em).
{code}
// find_value_last
field_value_last = line_e - 1;
while ((field_value_last >= field_value_first) &&
ParseRules::is_wslfcr(*field_value_last)) {
--field_value_last;
}
{code}
It's possible 3) should be its own Jira, but seems easy to do all this as one
fix honestly.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
