Commits
-------
3df5ec3 [HttpKernel] Add support for 'upload_max_filesize' ini directive in the Client
Discussion
----------
[HttpKernel] Add support for 'upload_max_filesize' ini directive
[HttpKernel] Add support for 'upload_max_filesize' ini directive in the Client
__This PR depends on #1542__
This PR prevent the SW Client from uploading files larger than the limit set in php.ini to closer mimic a real browser usage.
If both PR eventually gets merge `static protected function getMaxUploadFilesize()` should probably be factorized to the UploadedFile class.
---------------------------------------------------------------------------
by stloyd at 2011/07/05 13:35:06 -0700
+1 for both, I just have found similar "wtf" issues with "empty" `upload_max_filesize`.
---------------------------------------------------------------------------
by oscarballadares at 2011/07/05 15:13:23 -0700
I have opened an issue related to UPLOAD_ERR_INI_SIZE. There was no way to handle this exception.
Can you confirm please?
If this is the case I will close the issue I opened.
---------------------------------------------------------------------------
by vicb at 2011/07/05 23:04:08 -0700
@oscarballadares the PR you are looking for is most probably #1542 - which you should see in the message thread of your submitted issue.
The best would be for you to verify that PR #1542 fixes your issue and provide some feedback so that the issue can be close but only when the PR gets merged (if it fixes the issue).
Commits
-------
d58ba34 [Validator] Consider the ini directive 'upload_max_filesize' while validating an uploaded file (fixes GH-1441)
Discussion
----------
[Validator] FileValidator support for uploaded files
[Validator] Consider the ini directive 'upload_max_filesize' while validating an uploaded file (fixes GH-1441)
Added validator messages should get translated in all the available languages.
Commits
-------
03fee4f Fix permissions
431460f [Form] Remove choice or choice_list requirement as the following conditions already check enough and this condition prevents empty select forms (populated by ajax for example)
Discussion
----------
[Form] Choice fix
[Form] Remove choice or choice_list requirement as the following conditions already check enough and this condition prevents empty select forms (populated by ajax for example)
---------------------------------------------------------------------------
by stloyd at 2011/07/05 06:26:36 -0700
You should revert permission changes.
---------------------------------------------------------------------------
by fabpot at 2011/07/05 06:28:14 -0700
Why not replacing `if (!$options['choices'] && !$options['choice_list']) {` by `if (!isset($options['choices']) && !isset($options['choice_list'])) { `?
---------------------------------------------------------------------------
by beberlei at 2011/07/05 06:35:50 -0700
gnaa permission changes, i cant seem to configure my machine such that it does not do it, i have to do this on a per repository basis, very annoying.
@fabpot isset() is already guaranteed because these two options are in the defaults.
---------------------------------------------------------------------------
by beberlei at 2011/07/05 06:39:43 -0700
Fixed the permissions
---------------------------------------------------------------------------
by stof at 2011/07/05 06:48:37 -0700
@beberlei Can't you fix it in the global git config ?
---------------------------------------------------------------------------
by webda2l at 2011/07/05 09:48:58 -0700
I met the same problem this afternoon and vote for the isset solution. Better than nothing and work for me.
https://github.com/symfony/symfony/pull/1539
---------------------------------------------------------------------------
by stof at 2011/07/05 09:50:09 -0700
@webda2l why is a check that always return true better than nothing ? It adds overhead without adding any value in the code.
Commits
-------
3917ed7 Revert "* DateType, DateTimeType, TimeType: - a bit changed readability"
c85b815 Fixed few issues with Date and Time:
Discussion
----------
[Form] Fixed few issues with Date and Time
Fixed few issues with Date and Time:
* TimeType:
- seconds are no longer populated if "with_seconds" = false
- "widget = text" is now properly rendered (closes#1480)
* DateTimeToStringTransformer:
- fixed using not default "format" (probably fix#1183)
* DateType, DateTimeType, TimeType:
- fixed "input = datetime" and test covered
Commits
-------
164aea4 [Security] Add tests for the channel listener
d51cbc0 [Security] Remove useless attribute in basic authentication listener & test it
91e6dc9 [Security] Add tests for the anonymous authentication listener
3c2affb [Security] Update access listener constructor's prototype and add tests
81afd77 [Security] Add tests for the firewall map
aa6ae33 [Security] Remove useless attribute & var in firewall
Discussion
----------
Test security
---------------------------------------------------------------------------
by lsmith77 at 2011/06/29 13:41:07 -0700
@schmittjoh is probably the person to review this change ..
Commits
-------
418d6a0 [Routing] Fix syntax error when dumping routes with single quotes in the requirements or pattern
2b5e22d [Routing] Fix ApacheDumper when a space appears in a default value
6c7f484 [Routing] Fix dumper so it doesn't print trailing whitespace
761724a [Routing] Adjust urlescaping rules, fixes#752
Discussion
----------
[Router] Bunch o' Fixes
The first commit changes the escaping rule to fix issues I had previously, and #752 as well, here's from the full commit message:
Only + and % are now encoded in generated routes, since they are the only characters that, if not encoded, could cause problems/conflicts when decoded. Namely + turns into a space, and % followed by numbers could do funky things.
The matcher decodes everything which works since nothing will have %NN without being escaped, and + are escaped as well.
Second commit is just a test fix for the first
Third and fourth are simply dumper escaping issues, nothing to argue about.
Note that all changes have had test cases added, and I spent a few hours torturing/testing all this stuff with both Apache and PHP dumpers, in many browsers, and with URLs as wonky as `/%01%02%03%04%05%06%07%08%0E%0F%10%11%12%13%14%15%16%17%18%19%1A%1B%1C%1D%1E%1F%20!%22$%25&%27%28%29*+,-./0123456789:;%3C=%3E@ABCDEFGHIJKLMNOPQRSTUVWXYZ%5B%5D%5E_%60abcdefghijklmnopqrstuvwxyz%7B|%7D~/baz` which essentially represent the 1-255 char range minus ? and #.
The only issues I really encountered after all the patches were applied is that Apache refuses to match `%22` (= `"`) and `*` in a url. I guess it's just because they're not allowed chars in windows paths, but | and < > works fine though. Anyway this works with the PHP dumper, and it didn't work either without my patches so it's not like I broke it, I'm just saying for the record.
Commits
-------
1cc1027 added @Annotation to UniqueEntity
ee22c5d added a note to update file
efcb435 updated to doctrine changes
Discussion
----------
updated to doctrine changes
---------------------------------------------------------------------------
by excelwebzone at 2011/06/30 06:29:23 -0700
Should also be implemented to the Route class and to all SensioFrameworkExtraBundle annotation classes
* TimeType:
- seconds are no longer populated if "with_seconds" = false
- "widget = text" is now properly rendered (closes#1480)
* DateTimeToStringTransformer:
- fixed using not default "format" (probably fix#1183)
* DateType, DateTimeType, TimeType:
- fixed "input = datetime" and test covered
- a bit changed readability
Only + and % are now encoded in generated routes, since they are the only characters that, if not encoded, could cause problems/conflicts when decoded. Namely + turns into a space, and % followed by numbers could do funky things.
The matcher decodes everything which works since nothing will have %NN without being escaped, and + are escaped as well.
Commits
-------
f4c7333 Fix populating seconds when option "with_seconds" is set to false
Discussion
----------
[Form][DateTimeType] Fix invalid data when "with_seconds" = false
Fix populating seconds when option `with_seconds` is set to `false`.
Commits
-------
4e3406d Sync with master and clean up
ad5d2c1 Added to `TimeType` extension possibility to render form as `single_text` (similar to DateType option) (issue #1205) Adjusted `DateTimeType` to allow usage of this new feature
Discussion
----------
[Form][TimeType] Added possibility to render form as "single_text"
Added to `TimeType` extension possibility to render form as `single_text` (similar to `DateType` option) (issue #1205)
Adjusted `DateTimeType` to allow usage of this new feature
---------------------------------------------------------------------------
by ouardisoft at 2011/06/17 03:41:18 -0700
+1
---------------------------------------------------------------------------
by stloyd at 2011/06/21 01:05:51 -0700
@fabpot Any decision about this one ? I'm asking because I also have similar fix for #1323 but it requires this one ;-)
---------------------------------------------------------------------------
by fabpot at 2011/06/22 23:32:08 -0700
@stloyd: Can you rebase to master?
---------------------------------------------------------------------------
by stloyd at 2011/06/23 05:03:44 -0700
@fabpot Done.
Commits
-------
7db0b95 [HttpKernel] Removed unnecessary strtoupper
0891c57 [HttpKernel] Added test
1350645 [HttpKernel] Uppercased a few http methods
05c9906 [HttpKernel] Suppress response content for 304 responses out of the cache
Discussion
----------
HttpCache changes for 304 responses
Fixes#1413
Commits
-------
2d29a82 New test for Process, testing stdout and stderr at different stream sizes
Discussion
----------
Make run() fully non-blocking and fix potential other problems
Multiple changes:
1) make writing to process non-blocking too - otherwise there might be increased possibility for buffer deadlock
given big enough input data. Also now it's guaranteed that all stdin data will be written.
2) get rid of fgets() - fgets() isn't really good function to use in case of non-blocking sockets. Data loss possible.
---------------------------------------------------------------------------
by fabpot at 2011/06/22 07:11:55 -0700
Does it make https://github.com/symfony/symfony/pull/1365 obsolete?
---------------------------------------------------------------------------
by lenar at 2011/06/22 14:08:14 -0700
@fabpot: After reading, I really don't know. Let's hope. But ...
I now improved Process tests a bit to test stdout, stderr with different stream sizes and different
behaviours of child processes. Added it to non-blocking-process branch, commit 2d29a82412.
In my case, nothing fails, but maybe this helps other people. Or Windows people - I myself cannot test on Windows.
---------------------------------------------------------------------------
by fabpot at 2011/06/22 22:59:55 -0700
These tests pass on my Linux box but fail on my Mac.
---------------------------------------------------------------------------
by fabpot at 2011/06/22 23:05:14 -0700
Actually, on the Mac, the tests behave correctly but the exit code is `-1` instead of `0`.
---------------------------------------------------------------------------
by lenar at 2011/06/23 01:23:51 -0700
Could you check if the $this->status['running'] (after call to proc_get_status()) is true in the case you get -1.
On my linux I got it -1 couple of times. 99% of time it doesn't happen. I theorized it's because sometimes the child
process isn't finished enough and finally I got confirmation too that in case of -1 the process is still running (stats['running'] === true).
But it's really almost unreproducible on my Linux. So if you have this value every time it might be easier for you to find solution.
What comes into my mind:
1) maybe we should poll, let's say if process is still running we usleep(1000) and the try proc_get_status() again until not running. Maybe up to a 1 sec.
2) maybe, if the process is still running we can trust the return value subsequently given by proc_close()?
Or maybe there's some other problem on Mac.