Since I'll be out most of today, I'll just post the patch first; it'd
be great if it could also be applied to the 5_2 branch of course, but
that's not my call ;-)
As for compatibility, by default the patch would not change anything
to the current 5.2 behaviour, so I'm not sure how this poses a
problem. In addition, the documentation already states the
"ignore_errors" settings without mentioning the version it was
introduced in; a simple "introduced in 5.2.10" would work, no?
Reference documentation: http://sg.php.net/manual/en/context.http.php
Best,
Tjerk
On Sun, May 17, 2009 at 8:42 AM, Tjerk Anne Meesters <datib...@php.net> wrote:
> But ignore_errors is not used in 5.2 right? ;-) I also have a patch
> for that actually if you're interested =D
>
> On 5/17/09, Arnaud Le Blanc <lbarn...@php.net> wrote:
>> Yes, the fix requires "ignore_errors", which is new in 5.3. The bug
>> can't be fixed in 5.2 without breaking compatibility.
>>
>> Regards,
>>
>> Arnaud
>>
>> On Sun, 2009-05-17 at 02:02 +0300, Jani Taskinen wrote:
>>> Last 7 commits or so, none went to PHP_5_2. Intentionally?
>>> We usually fix bugs in the branch they're reported for also..
>>>
>>> --Jani
>>>
>>>
>>> Arnaud Le Blanc kirjoitti:
>>> > I commited your change as it fixes a bug, thanks for the patch :)
>>> >
>>> > I also followed your idea of a server to test http streams (ftp
>>> > extension does this, too) and wrote a test for this bug. Other http
>>> > tests are welcome if you want to write some.
>>> >
>>> > Regards,
>>> >
>>> > Arnaud
>>> >
>>> > On Sat, 2009-05-16 at 16:42 +0800, Tjerk Anne Meesters wrote:
>>> >> On a related note, I've been working on a phpt to verify the behaviour
>>> >> during which I've noticed a lack of test cases for networked functions
>>> >> in general.
>>> >>
>>> >> The solution I have in mind depends on pcntl_fork(), which practically
>>> >> takes out Windows test targets. It essentially forks out a child to
>>> >> perform the test while the parent starts a socket_server.
>>> >>
>>> >> I suppose it would be nice if the test framework could somehow be
>>> >> extended to pose as a webserver to verify typical http behaviour
>>> >> inside stream wrappers =D
>>> >>
>>> >> This can't be the first time that this has been a point of discussion;
>>> >> my apologies if this has been outright rejected before ;-)
>>> >>
>>> >> On 5/16/09, Arnaud Le Blanc <lbarn...@php.net> wrote:
>>> >>> On Fri, 2009-05-15 at 11:49 -0400, Ilia Alshanetsky wrote:
>>> >>>> the patch looks good.
>>> >>> Same here. Lukas, Johannes, can this be commited to 5.3 ?
>>> >>>
>>> >>> Regards,
>>> >>>
>>> >>> Arnaud
>>> >>>
>>> >>>
>>> >>>>
>>> >>>> Ilia Alshanetsky
>>> >>>>
>>> >>>>
>>> >>>>
>>> >>>>
>>> >>>> On 15-May-09, at 11:36 AM, Tjerk Anne Meesters wrote:
>>> >>>>
>>> >>>>> Hi Arnaud,
>>> >>>>>
>>> >>>>> Thanks for the tip, please find patch attached.
>>> >>>>>
>>> >>>>>
>>> >>>>> Best,
>>> >>>>> Tjerk
>>> >>>>>
>>> >>>>> On Fri, May 15, 2009 at 10:58 PM, Arnaud Le Blanc <lbarn...@php.net>
>>> >>>>> wrote:
>>> >>>>>> Hi,
>>> >>>>>>
>>> >>>>>> On Fri, 2009-05-15 at 22:16 +0800, Tjerk Anne Meesters wrote:
>>> >>>>>>> Hi,
>>> >>>>>>>
>>> >>>>>>> I've been extending the pecl/oauth code to work with php stream
>>> >>>>>>> wrappers in cases whereby curl is not enabled, but I've hit a
>>> >>>>>>> snag.
>>> >>>>>>>
>>> >>>>>>> The documentation states that enabling the "ignore_errors" flag
>>> >>>>>>> will
>>> >>>>>>> fetch the content even on failure status codes. At the same time,
>>> >>>>>>> setting "max_redirects" to <= 1 indicates that no redirects will
>>> >>>>>>> be
>>> >>>>>>> followed.
>>> >>>>>>>
>>> >>>>>>> It does not state that setting "max_redirects" <= 1 would actually
>>> >>>>>>> trigger an error, though in the code it returns (php_stream *)NULL
>>> >>>>>>> making it impossible to even get the headers out of the
>>> >>>>>>> wrapperdata
>>> >>>>>>> hash unless STREAM_ONLY_GET_HEADERS is used.
>>> >>>>>>>
>>> >>>>>>> So, either setting "max_redirects" <= 1 should not throw NULL back
>>> >>>>>>> into the caller code or "ignore_error" behaviour should be
>>> >>>>>>> extended to
>>> >>>>>>> prevent that from happening.
>>> >>>>>>>
>>> >>>>>>> My attached patch favours to fix the latter, since the
>>> >>>>>>> "ignore_errors"
>>> >>>>>>> is a relatively new flag. Hope it will be accepted for the 5.3
>>> >>>>>>> release.
>>> >>>>>> max_redirects's purpose is to avoid infinite loops, etc. Stream
>>> >>>>>> functions are expected to return an error in case the limit is
>>> >>>>>> exceeded.
>>> >>>>>> If they do not, scripts will get an empty body without noticing the
>>> >>>>>> error (changing this may break existing scripts).
>>> >>>>>>
>>> >>>>>> ignore_errors is a new flag, it forces scripts to check HTTP status
>>> >>>>>> code
>>> >>>>>> (so that scripts will notice non-200 ones) and changing it to
>>> >>>>>> ignore
>>> >>>>>> redirects when max_redirects is exceeded seems to be a good
>>> >>>>>> solution.
>>> >>>>>>
>>> >>>>>>> (not sure whether attachments will be allowed ... give me a ping
>>> >>>>>>> if it
>>> >>>>>>> doesn't make it through)
>>> >>>>>> The list accepts only text/plain attachments (try renaming your
>>> >>>>>> patch
>>> >>>>>> to .txt).
>>> >>>>>>
>>> >>>>>>
>>> >>>>>> Regards,
>>> >>>>>>
>>> >>>>>> Arnaud
>>> >>>>>>
>>> >>>>>>
>>> >>>>>>
>>> >>>>>
>>> >>>>>
>>> >>>>> --
>>> >>>>> --
>>> >>>>> Tjerk
>>> >>>>> <http_fopen_wrapper.c.patch.txt>--
>>> >>>>> PHP Internals - PHP Runtime Development Mailing List
>>> >>>>> To unsubscribe, visit: http://www.php.net/unsub.php
>>> >>>
>>> >>
>>> >
>>> >
>>>
>>
>>
>
>
> --
> --
> Tjerk
>
--
--
Tjerk
Index: http_fopen_wrapper.c
===================================================================
RCS file: /repository/php-src/ext/standard/http_fopen_wrapper.c,v
retrieving revision 1.99.2.12.2.17
diff -u -r1.99.2.12.2.17 http_fopen_wrapper.c
--- http_fopen_wrapper.c 14 May 2009 16:15:38 -0000 1.99.2.12.2.17
+++ http_fopen_wrapper.c 17 May 2009 03:42:44 -0000
@@ -104,7 +104,7 @@
size_t chunk_size = 0, file_size = 0;
int eol_detect = 0;
char *transport_string, *errstr = NULL;
- int transport_len, have_header = 0, request_fulluri = 0;
+ int transport_len, have_header = 0, request_fulluri = 0, ignore_errors
= 0;
char *protocol_version = NULL;
int protocol_version_len = 3; /* Default: "1.0" */
struct timeval timeout;
@@ -552,8 +552,12 @@
} else {
response_code = 0;
}
+ if (context &&
SUCCESS==php_stream_context_get_option(context, "http", "ignore_errors",
&tmpzval)) {
+ ignore_errors = zend_is_true(*tmpzval);
+ }
+
/* when we request only the header, don't fail even on
error codes */
- if (options & STREAM_ONLY_GET_HEADERS) {
+ if ((options & STREAM_ONLY_GET_HEADERS) ||
ignore_errors) {
reqok = 1;
}
switch(response_code) {
@@ -631,7 +635,7 @@
}
if (!reqok || location[0] != '\0') {
- if (options & STREAM_ONLY_GET_HEADERS && redirect_max <= 1) {
+ if (((options & STREAM_ONLY_GET_HEADERS) || ignore_errors) &&
redirect_max <= 1) {
goto out;
}
--
PHP Internals - PHP Runtime Development Mailing List
To unsubscribe, visit: http://www.php.net/unsub.php
