apr-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "William A. Rowe, Jr." <wr...@rowe-clan.net>
Subject Re: [PATCH] fix design bug in apr_file_gets()
Date Mon, 28 Apr 2003 16:42:00 GMT
At 01:13 PM 4/27/2003, Jeff Trawick wrote:
>change apr_file_gets() to return APR_SUCCESS if returning any data
>
>this change will resolve bugs in such callers as mod_cgi and mod_cgid which did not have
logic to find data with no final '\n' at end of file; conceivably it will make a number of
other callers accept data when there is no '\n'
>
>if somebody really really wants until apr 1.0 to fix this, we can have apr_file_gets_foo()
until then for callers which would rather have apr do the heavy lifting instead of adding
missing logic to deal with the fact that on normal files (pipes) where there are no I/O errors
good data can come even when rv != APR_SUCCESS

No need to wait.  Consider that anyone using apr_file_gets() already expects
to have an EOF.  The question is if they are prepared to read data.

As long as we give them no data I believe the change won't break any
applications.  So ++1 since this is an easy app bug we can help to
alleviate.

The only risk is users who might use data+EOF to check for missing \n.
This would be a silly side-effect dependency, but it belongs in CHANGES
just in case anyone had done so.

Bill



Mime
View raw message