This PR was merged into the 2.0 branch.

Commits
-------

880da01 [Process] In edge cases `getcwd()` can return `false`, then `proc_open()` should get `null` to use default value (the working dir of the current PHP process)

Discussion
----------

[2.0][Process] In edge cases `getcwd()` can return `false`

Bug fix: yes
Feature addition: no
BC break: no
Symfony2 tests pass: yes

In edge cases `getcwd()` can return `false`, then `proc_open()` should get `null` to use default value (the working dir of the current PHP process).

---------------------------------------------------------------------------

by stloyd at 2013-01-08T12:43:40Z

I guess that this could be related to #6496, as error code `267` at Windows means:
[`ERROR_DIRECTORY - The directory name is invalid.`](http://msdn.microsoft.com/en-us/library/ms681382%28v=vs.85%29.aspx#error_directory)

---------------------------------------------------------------------------

by Seldaek at 2013-01-08T12:57:38Z

If null already uses the current working directory, what's the point of calling getcwd() at all?

---------------------------------------------------------------------------

by stloyd at 2013-01-08T13:03:06Z

@Seldaek TBH. I don't have idea =) It seems that it's there _from the beginning_, but yeah, I was a bit confused by usage of it too...

---------------------------------------------------------------------------

by fabpot at 2013-01-09T08:13:24Z

What about removing the code altogether?

---------------------------------------------------------------------------

by stloyd at 2013-01-09T08:22:55Z

@fabpot I'm ok with that, just not sure it will not be an BC break...

---------------------------------------------------------------------------

by Seldaek at 2013-01-09T08:24:57Z

php.net says `or NULL if you want to use the default value (the working
dir of the current PHP process)` which sounds like getcwd() to me.

---------------------------------------------------------------------------

by Seldaek at 2013-01-09T08:26:32Z

For full BC though, `getWorkingDirectory` should `return $this->cwd ?:
getcwd();` Then at least if that call fails the whole process isn't
failing. I don't see why anyone would use that getter though.

---------------------------------------------------------------------------

by stloyd at 2013-01-10T12:43:59Z

@fabpot @Seldaek What do you think about this now?

---------------------------------------------------------------------------

by Seldaek at 2013-01-10T12:58:39Z

:+1:

…ymfony#7082

This PR was merged into the 2.0 branch.

Commits
-------

179cd58 [Process] Fix regression introduced in #6620 / 880da01, fixes #7082

Discussion
----------

[Process][2.0] getcwd failure fix

Fix regression introduced in #6620
Fixes #7082

For reference, here is the current behavior I saw:

PHP 5.4.11, windows:

```
5.4.11\php.exe -r "chdir('c:\\'); var_dump(getcwd()); $p = proc_open('pwd', [['pipe', 'r'], ['pipe', 'w'], ['pipe', 'w']], $pipes, null); var_dump(stream_get_contents($pipes[1]));"
string(3) "C:\\"
string(14) "/c/Users/seld\n"
```

(I use pwd which is a unix util so it dumps a funny path, but don't look at that)

PHP 5.5alpha4, windows (seems fixed):

```
5.5.0a4\php.exe  -r "chdir('c:\\'); var_dump(getcwd()); $p = proc_open('pwd', [['pipe', 'r'], ['pipe', 'w'], ['pipe', 'w']], $pipes, null); var_dump(stream_get_contents($pipes[1]));"
string(3) "C:\"
string(3) "/c\n"
```

PHP 5.3.10, ubuntu:

```
php -r "chdir('/'); var_dump(getcwd()); \$p = proc_open('pwd', array(array('pipe', 'r'), array('p
ipe', 'w'), array('pipe', 'w')), \$pipes, null); var_dump(stream_get_contents(\$pipes[1]));"
string(1) "/"
string(2) "/\n"
```

Since the permission issue that #6620 originally was fixing is most likely not gonna happen on windows, this seems like a safe enough compromise. Ideally a check for PHP<5.5 should be introduced, but I would like to be sure it's been fixed and is not just a lucky coincidence (/cc @pierrejoye)

I would recommend merging fast and maybe adding the version check later, since it breaks composer create-project on windows. The workaround being: cd in the dir and run `composer install` again to finalize the project setup.

---------------------------------------------------------------------------

by vicb at 2013-02-17T20:12:17Z

Thanks @Seldaek !

Would you mind creating an issue for the version check so that it doesn't get lost ?

---------------------------------------------------------------------------

by Seldaek at 2013-02-18T10:43:56Z

@vicb done.

---------------------------------------------------------------------------

by vicb at 2013-02-18T11:45:16Z

thanks !

* 2.0:
  [Process] Fix regression introduced in #6620 / 880da01, fixes #7082

* 2.1:
  [Process] Fix regression introduced in #6620 / 880da01, fixes #7082

* 2.2: (22 commits)
  [Process] Fix regression introduced in #6620 / 880da01, fixes #7082
  [HttpKernel] added a unit for the previous commit (closes #7025)
  [HttpFoundation] fixed, overwritten CONTENT_TYPE
  [BrowserKit] fixed test added in the previous merge (refs #7059)
  [FrameworkBundle] tweaked reference dumper command (see #7093)
  Remove unnecessary comment and change test name
  [Config] tweaked dumper to indent multi-line info
  [HttpKernel] added some tests for previous merge
  Fix REMOTE_ADDR for cached subrequests
  [FrameworkBundle] CSRF should be on by default
  [WebProfilerBundle] removed dependency on FrameworkBundle (closes #6949)
  [HttpKernel] added error display suppression when using the ErrorHandler (if not, errors are displayed twice, refs #6254)
  [HttpFoundation] tweaked previous merge
  [HttpFoundation] Added getter for httpMethodParameterOverride state
  Create validators.lv.xlf
  [Process] Warn user with a useful message when tmpfile() failed
  [BrowserKit] added a test to make sure HTTP authentication is preserved when submitting a form
  Remove array type hint from GetResponseForControllerResultEvent::setControllerResult()
  bumped Symfony version to 2.2.0-DEV
  Revert "merged branch povilas/issue_6101 (PR #6708)"
  ...

* 2.2: (22 commits)
  [Process] Fix regression introduced in symfony#6620 / 880da01, fixes symfony#7082
  [HttpKernel] added a unit for the previous commit (closes symfony#7025)
  [HttpFoundation] fixed, overwritten CONTENT_TYPE
  [BrowserKit] fixed test added in the previous merge (refs symfony#7059)
  [FrameworkBundle] tweaked reference dumper command (see symfony#7093)
  Remove unnecessary comment and change test name
  [Config] tweaked dumper to indent multi-line info
  [HttpKernel] added some tests for previous merge
  Fix REMOTE_ADDR for cached subrequests
  [FrameworkBundle] CSRF should be on by default
  [WebProfilerBundle] removed dependency on FrameworkBundle (closes symfony#6949)
  [HttpKernel] added error display suppression when using the ErrorHandler (if not, errors are displayed twice, refs symfony#6254)
  [HttpFoundation] tweaked previous merge
  [HttpFoundation] Added getter for httpMethodParameterOverride state
  Create validators.lv.xlf
  [Process] Warn user with a useful message when tmpfile() failed
  [BrowserKit] added a test to make sure HTTP authentication is preserved when submitting a form
  Remove array type hint from GetResponseForControllerResultEvent::setControllerResult()
  bumped Symfony version to 2.2.0-DEV
  Revert "merged branch povilas/issue_6101 (PR symfony#6708)"
  ...