On 10/30/2015 06:47 AM, David Malcolm wrote:
The typename suggestion seems to be at least somewhat controversial,
whereas (I hope) the misspelled field names suggestion is more
acceptable.
Hence I'm focusing on the field name lookup for now; other uses of the
algorithm (e.g. the typename lookup) could be done in followup patches,
but I'm deferring them for now in the hope of getting the simplest case
into trunk as a first step. Similarly, for simplicity, I didn't
implement any attempt at error-recovery using the hint.
The following patch kit is in two parts (for ease of review; they would
be applied together):
patch 1: Implement Levenshtein distance
patch 2: C FE: suggest corrections for misspelled field names
I didn't implement a limiter, on the grounds that this only fires
once per "has no member named" error, and so is unlikely to slow
things down noticeably.
Successfully bootstrapped®rtested the combination of these two
on x86_64-pc-linux-gnu (adds 11 new PASS results to gcc.sum)
OK for trunk?
gcc/Makefile.in | 1 +
gcc/c/c-typeck.c | 70 +++++++++++-
gcc/spellcheck.c | 136 +++++++++++++++++++++++
gcc/spellcheck.h | 32 ++++++
gcc/testsuite/gcc.dg/plugin/levenshtein-test-1.c | 9 ++
gcc/testsuite/gcc.dg/plugin/levenshtein_plugin.c | 64 +++++++++++
gcc/testsuite/gcc.dg/plugin/plugin.exp | 1 +
gcc/testsuite/gcc.dg/spellcheck-fields.c | 63 +++++++++++
8 files changed, 375 insertions(+), 1 deletion(-)
create mode 100644 gcc/spellcheck.c
create mode 100644 gcc/spellcheck.h
create mode 100644 gcc/testsuite/gcc.dg/plugin/levenshtein-test-1.c
create mode 100644 gcc/testsuite/gcc.dg/plugin/levenshtein_plugin.c
create mode 100644 gcc/testsuite/gcc.dg/spellcheck-fields.c
I'm going to assume you got levenshtein's algorithm reasonably correct.
This is OK for the trunk. Obviously I'd like to see it extend into the
other front-ends (C++ in particular). Then I'd like to see it extend
beyond just misspelled field names.
jeff
