On Wed, 2011-07-27 at 10:32 +0200, Denys Duchier wrote:
Hi David, here is the new transcript after updating:
Thanks! Notes follow inline below:
tests/plugin/diagnostics: FAIL -tests/plugin/diagnostics/input.c: In function 'main': +tests/plugin/diagnostics/input.c: In function ‘main’:
It looks like it's emitting unicode quote characters: U+2018 LEFT SINGLE QUOTATION MARK and: U+2019 RIGHT SINGLE QUOTATION MARK whereas the test is expecting the plain ASCII: U+0027 APOSTROPHE
I tried to change this expectation in f42aa1507acbfc5b0274ff2b65875d5510899fcb http://git.fedorahosted.org/git/?p=gcc-python-plugin.git;a=commit;h=f42aa150...
where I started setting LANG=C when running the tests.
The message is emitted in 4.6.0 by gcc/langhooks.c: pp_printf (context->printer, _("In function %qs"), identifier_to_locale (lang_hooks.decl_printable_name (fndecl, 2)));
"%qs" means a quoted string in gcc/prettyprint.c, and it seems to use: extern const char *open_quote; extern const char *close_quote; from gcc/intl.h, set up in gcc/intl.c
Rereading locale(1), it seems LANG is merely used for defaults, and LC_ALL overrides everything; I was able to reproduce the failure by running:
$ LC_ALL=en_US.UTF-8 make
so I now think I should be setting LC_ALL=C instead of LANG=C when trying to compare against an expected output.
Fix committed as 4079ec5a6b6ef2e012c0f14a07e4bc0b8c5530b1 ; with that, all tests run to completion for me, for the above invocation of "make".
tests/plugin/options: FAIL
- option.is_enabled: False
- option.is_enabled: True
This is the minor failure we saw earlier in this discussion. I've tried to fix this in b08a84c493eb9ca569652aa208a3572e5a8fe02d
tests/cpychecker/PyArg_ParseTuple/keywords: FAIL tests/cpychecker/PyArg_ParseTuple/incorrect_codes_S_and_U: FAIL tests/cpychecker/PyArg_ParseTuple/incorrect_code_z_hash: FAIL tests/cpychecker/refcounts/incorrect_py_none: FAIL tests/cpychecker/refcounts/use_after_dealloc: FAIL tests/cpychecker/refcounts/too_many_increfs: FAIL tests/cpychecker/refcounts/not_setting_exception: FAIL tests/cpychecker/refcounts/returning_dead_object: FAIL tests/cpychecker/refcounts/missing_decref: FAIL tests/cpychecker/refcounts/object_leak: FAIL
These remaining failures appear to all be the same locale issue as above, so (I hope) are now fixed.
If you pull the latest version from git, how are things looking?
Dave