On 08/04/2016 08:27 AM, David Malcolm wrote:
As for test coverage, v2 and v3 of the kit add over a thousand lines of
selftest code that heavily exercise string lexing, using the
line_table_case machinery to run the tests with various interesting
boundary conditions with line_table (e.g. near
LINE_MAP_MAX_LOCATION_WITH_PACKED_RANGES).
In terms of test coverage of the fallbacks, patch 2 of v3 of the kit
directly exercises the substr_loc.get_range in
gcc.dg/plugin/diagnostic_plugin_test_string_literals.c via
gcc.dg/plugin/diagnostic-test-string-literals-1.c, and some of the
tests there cover the failures, via:
error_at (strloc, "unable to read substring range: %s", err);
which we wouldn't do in a normal diagnostic (but which is appropriate
for testing the machinery itself).
Patch 3 of the v3 kit adds a format_warning_va function to c-format.c
which is responsible for dealing with failures:
https://gcc.gnu.org/ml/gcc-patches/2016-08/msg00204.html
THanks for pointing this out. I hadn't started looking at the meat of
the on-demand locations until this morning.
Looking at patch 3, there's a fair amount of end-to-end testing in
gcc.dg/format/diagnostic-ranges.c but it looks like I forgot to add an
end-to-end test there of failure due to stringification; I can add one.
Is the rest of the v3 patch kit reviewable?
Absolutely. I wasn't trying to imply that it wasn't -- in fact most of
it is self-approvable stuff and I've only got a couple questions about
the rest.
Jeff
