report_webkit not thread-safe: risk of document corruption

Copy of hot fix for memory in order to do the MP on OCB and official

=== modified file 'report_webkit/webkit_report.py'
--- report_webkit/webkit_report.py 2013-10-27 12:31:04 +0000
+++ report_webkit/webkit_report.py 2014-03-11 11:44:49 +0000
@@ -35,6 +35,7 @@
 import sys
 from openerp import report
 import tempfile
+import thread
 import time
 import logging

@@ -157,6 +158,7 @@
             webkit_header = report_xml.webkit_header
         tmp_dir = tempfile.gettempdir()
         out_filename = tempfile.mktemp(suffix=".pdf", prefix="webkit.tmp.")
+ timestamp = '%x.%x' % (thread.get_ident(), 1e7 * time.time())
         file_to_del = [out_filename]
         if comm_path:
             command = [comm_path]
@@ -169,7 +171,7 @@
         if header :
             head_file = file( os.path.join(
                                   tmp_dir,
- str(time.time()) + '.head.html'
+ timestamp + '.head.html'
                                  ),
                                 'w'
                             )
@@ -180,7 +182,7 @@
         if footer :
             foot_file = file( os.path.join(
                                   tmp_dir,
- str(time.time()) + '.foot.html'
+ timestamp + '.foot.html'
                                  ),
                                 'w'
                             )
@@ -203,7 +205,7 @@
             command.extend(['--page-size', str(webkit_header.format).replace(',', '.')])
         count = 0
         for html in html_list :
- html_file = file(os.path.join(tmp_dir, str(time.time()) + str(count) +'.body.html'), 'w')
+ html_file = file(os.path.join(tmp_dir, timestamp + str(count) +'.body.html'), 'w')
             count += 1
             html_file.write(html.encode('utf-8'))
             html_file.close()

I've just discovered something worse: "self.parser_instance" is shared between threads.

And this shared "parser_instance" has important attributes :

 * self.parser_instance.cr: which means that the same PostgreSQL cursor is accessed from more than one thread : it breaks randomly

 * self.parser_instance.localcontext['objects'] : which means that it might explain that the wrong document was generated (the issue reported here, on bug #1290820)

These is really a can of worms when using "report_webkit" in a Production environment.

We have a local patch for it, I'll do a second merge proposal soon.

I confirm this bug as I had the same issue.
As I don't see your second MP I'll create a other MP to solve the issue with self.parser_instance

About file names, I prefered to use NamedTemporaryFiles
https://code.launchpad.net/~camptocamp/openobject-addons/7.0-fix-1319095/+merge/219399

Here is the second MP:

https://code.launchpad.net/~camptocamp/openobject-addons/7.0-fix-1319109/+merge/219406