http_fopen_wrapper.c - Handle HTTP headers with varying white space

[IMSoP]

The stream handler assumed all HTTP headers contained exactly one space,
but the standard says there may be zero or more. Should fix Bug #47021,
and any other edge cases caused by a web server sending unusual spacing,
e.g. the MIME type discovered from Content-Type: can no longer contain
leading whitespace.

[nikic]

Declarations need to go at the top of the block in C89.

[IMSoP]

Thanks for looking over this so quickly! I'll hopefully have time to look into your feedback properly tomorrow, and update the patch. :)

[hlarsen]

hey all,

i'd just like to check the status of this - it looks like we're seeing an issue that this patch will solve.

[IMSoP]

@nikic Since you reviewed before, does this look OK now? I'm pretty sure the Travis error was just a false report.

[krakjoe]

awaits nikitas response

[nikic]

Question: What should happen with the headers we add to the $http_response_headers variable? As the PR implements it right now, if there is trailing whitespace, the header would contain a NUL byte. That's obviously not right, but I'm not sure whether its better to a) provide the header as-is (with trailing whitespace) or b) strip trailing whitespace.

The latter would certainly be easier to implement. Some of the APIs used there don't accept explicit lengths :/

[krakjoe]

The spec indicates that the trailing white space is allowed, however since it serves no practical purpose, I see no harm in removing it - the vast majority of the time, the programmer is surely going to do that anyway, and if they don't their code may exhibit unexpected behaviour.

[nikic]

Modified version merged as 5146d9f with a followup in ec1b7b9. Trailing whitespace is stripped from headers in $http_response_header. A "header" that only consists of whitespace is not considered the start of the body.

Please double check if the merged patch is fine.