Age | Commit message (Collapse) | Author |
|
connections, please have a look at the log files for:
"server_response: request %d in %s not finished"
where %d is the request number in a persistent connection and %s is
either server_file or server_fcgi.
The server_fcgi.c code also includes "//"-commented code for two
possible solutions: either reset after each FCGI_END_REQUEST or after
finishing the end request or any possible padding.
|
|
|
|
|
|
|
|
Change httpd(8) to use C99-style fixed-width integers (uintN_t instead
of u_intN_t) and replace u_int with unsigned int. Mixing both
variants is a bad style and most contributors seem to prefer this
style; it also helps us to get used to it, portability, and
standardization.
Theoretically no binary change, except one in practice: httpd.o has a
different checksum because gcc with -O2 pads/optimizes "struct
privsep" differently when using "unsigned int" instead "u_int" for the
affected members. "u_int" is just a typedef of "unsigned int", -O0
doesn't build the difference and clang with -O2 doesn't do it either -
it is just another curiosity from gcc-land.
OK semarie@
|
|
|
|
lpereira-str-find-aux-neg
|
|
|
|
It might return the return value of match_error(), which returns
(-1).
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
From Leandro Pereira (@lafp on twitter)
|
|
|
|
- add a counter for limiting the search for repetitor items ('*', '+',
'-' and '?')
- add test case for this new kind of error
|
|
- revert "normal" call to optimized-tail-call-recursion for '*', '?' and
'-'
- add a "pattern too complex" test to !singlematch() branch, so that it
is limited on call numbers.
|
|
- switch '*', '?' and '-' to normal call, in order to pass into "too
complex" check
- remove assert()
- on error, quit early
|
|
|
|
|
|
After the last change, we also have to url_encode $SERVER_NAME and
$REMOTE_USER before using them in the Location.
From Sebastien Marie (semarie)
|
|
- there is no need to escape "%" with "%%" in error message
- corrects a comment
- add a check for error
|
|
|
|
|
|
Discussed with Sebastien Marie
|
|
an error could be setted in classend(). check and propagate it if any.
|
|
different semantics as we cannot just abort as Lua does. So we have
to check return values carefully and I missed a few in the transition.
Patch by Sebastien Marie
|
|
a panic when using a ")(" pattern.
Found by Sebastien Marie
|
|
implementation.
|
|
user input and not the constants from the configuration. This makes
it possible to specify chars like '?' and '=' in the uri.
|
|
|
|
according to RFC 3875 PATH_INFO should either contain a
full path or be empty.
|
|
|
|
|
|
|
|
|