Shows in buildbots:

/home/dje/cpython-buildarea/3.x.edelsohn-sles-z/build/Modules/_json.c: In function ‘scanner_new’:
/home/dje/cpython-buildarea/3.x.edelsohn-sles-z/build/Modules/_json.c:1212:5: warning: comparison is always false due to limited range of data type [-Wtype-limits]
     if (s->strict < 0)
     ^

This means that the error check doesn't work right.

New changeset 541bd28941af407f2329e6a540d6367e5625b115 by Serhiy Storchaka (Segev Finer) in branch 'master':
bpo-30911: Fix a crash in json on platforms with unsigned char (#2684)
https://github.com/python/cpython/commit/541bd28941af407f2329e6a540d6367e5625b115

Thank you for noticing this warning Segev.

Actually the issue was more serious. This caused a crash when pass the strict argument that raises an exception when converted to bool. PR 2690 adds test for this (actually it just ports recently added tests from simplejson).

New changeset d3aaa2f1496aae0809c9ec9623fa528d3a2c16c2 by Serhiy Storchaka in branch 'master':
bpo-30911: Add tests for bad boolean arguments for accelerated json (#2690)
https://github.com/python/cpython/commit/d3aaa2f1496aae0809c9ec9623fa528d3a2c16c2

> Actually the issue was more serious.

If the bug is serious, it deserves a NEWS entry, no? Moreover, the fix+test should be backported to 2.7, 3.5 and 3.6, no?

Introduced in https://github.com/python/cpython/commit/ac5bbd43bc7b769c13ae0412cb28a3521f4d4ff1 which is only in 3.7 that means this wasn't ever released. I can send a PR with a NEWS entry if it still needs one?

Oh ok. If it's a regression in the master branch, the code was never released with the bug. In this case, there is no need to document such fix. Thank you :-) I close the issue.

Thanks for the fix Sergev, thanks for the test Serhiy ;-)

New changeset 7d0a99582050ccac2e7126484d23072be342f317 by Serhiy Storchaka in branch '3.6':
[3.6] bpo-30911: Add tests for bad boolean arguments for accelerated json (GH-2690) (#2692)
https://github.com/python/cpython/commit/7d0a99582050ccac2e7126484d23072be342f317

The tests are backported to other branches. This is a one of reasons why I created a separate PR for tests.

New changeset 2ec2c5f0b05efdcb7919c477591f1cb48bfb0aa9 by Serhiy Storchaka in branch '3.5':
[3.5] bpo-30911: Add tests for bad boolean arguments for accelerated json (GH-2690) (#2693)
https://github.com/python/cpython/commit/2ec2c5f0b05efdcb7919c477591f1cb48bfb0aa9

New changeset b6dae2e40d14436d77d3f0bfe6a5b4ae73c83c6e by Serhiy Storchaka in branch '2.7':
[2.7] bpo-30911: Add tests for bad boolean arguments for accelerated json (GH-2690) (#2694)
https://github.com/python/cpython/commit/b6dae2e40d14436d77d3f0bfe6a5b4ae73c83c6e

It is possible that the change introduced a reference leak. See for example:
http://buildbot.python.org/all/builders/AMD64%20Windows8.1%20Refleaks%203.5/builds/49

Serhiy's change added tests to test_json, so you can say this unearthed ref leaks rather than introduced them. Probably needs a separate bpo issue and fix.

Created http://bugs.python.org/issue30936

New changeset 3dbe11a1219d148e4a133449a86ac5d157e35d35 by Serhiy Storchaka in branch '2.7':
[2.7] bpo-30911: Add tests for bad boolean arguments for accelerated json (GH-2690) (#2729)
https://github.com/python/cpython/commit/3dbe11a1219d148e4a133449a86ac5d157e35d35