| Commit message (Collapse) | Author | Age |
| |
|
| |
|
| |
|
|
|
|
|
|
| |
many PHPDocs
some unused variables
some dynamically declared variables declared
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Sometimes users report bugs or problems releated to HTTP connections
made by our HTTPCient (eg. in the extension manager or XMLRPC).
Debugging those problems previously required modifying some sources to
enable the debugging output of the HTTPClient.
This patch adds a way to enable the debugging via a request parameter.
Since HTTP requests are often done only *after* some button was clicked,
the parameter is also looked for in the referrer of a request.
Debugging output is enabled through the existence of the parameter
'httpdebug' and is directly printed to the screen. Debugging HTTP only
works when allowdebug is enabled.
|
| |
|
| |
|
|
|
|
| |
the auth info was not working.
|
| |
|
| |
|
|
|
|
| |
if a needed CONNECT tunnel fails
|
| |
|
| |
|
|
|
|
| |
this ensures nested POST data is correctly encoded
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
|\
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
* master: (169 commits)
added PCRE UTF-8 checks to do=check FS#2636
avoid multiple paralell update checks
fix regression bug in HTTPClient FS#2621
changed PAGEUTILS_ID_HIDEPAGE to has BEFORE/AFTER
TarLib code cleanup
TarLib: fixed appending in non-dynamic mode
fixed third method of adding files in TarLib
fix lone zero block in TarLib created archives
fix use of constructor in TarLib
Slovak language update
Korean language update
Latvian language update
added event PAGEUTILS_ID_HIDEPAGE
added test for isHiddenPage()
removed redundant variables in tpl_include_page() (because of 3ff8773b)
added cut off points for mobile devices as parameters to style.ini
Corrected typo: ruke -> rule
Persian language update
Spanish language update
russian language update
...
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
In the recent refactoring of the HTTPClient, a problem with certain
systems was reintroduced. On these systems a select() call always
waits for a timeout on the first call before working properly on the
second call.
This patch reintroduces the shorter timeouts with usleep rate limiting
again.
Since this bug is not reproducible on other systems it can't be unit
tested unfortunately.
|
|/
|
|
| |
The included test cases currently expect a squid at localhost:3128
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
| |
Reading from a socket is handled by functions that encapsulate the error
handling and timeout conditions.
_readData reads blocks of data.
_readLine reads a single line.
|
|
|
|
|
| |
This is the first step in refactoring the socket reader to be more
resilient and easier to debug.
|
| |
|
| |
|
| |
|
| |
|
| |
|
|
|
|
| |
The previous regexp approach failed for empty headers.
|
|
|
|
|
| |
On certain environments, stream_select might produce temporary errors
when file descriptors are running scarce.
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
This changes the HTTP stream to blocking while writing to the stream
using select() to handle timeouts. Addtionally, wwriting is done in 4k
block now (as it is done with reading).
This is supposed to fix a problem with writing to a SSL stream that is
not quite ready.
Reading from the stream continues to be non-blocking as before.
|
|\ |
|
| | |
|
| |
| |
| |
| |
| | |
since we don't use setter/getters for the other options it doesn't make
sense to have them for the keep-alive function
|
| |
| |
| |
| |
| |
| |
| | |
As soon as something goes wrong while querying a http server
do not reuse the same connection again, its state is undefined.
In addition, check the connection for feof() before reusing it.
|
| |
| |
| |
| |
| |
| | |
Using a static context for the connection pool allows connection
reuse throughout the whole application without additional changes
in other places.
|