[ccan] CCAN Progress for October

Rusty Russell rusty at rustcorp.com.au
Fri Oct 8 17:03:54 EST 2010


Hi all,

   It's been over a month since my last progress report, so I thought I'd
send out another.

My favourite improvement has been that ccanlint now tries various heuristics
to compile example code fragments from comment headers.  All the examples in
the tree now compile successfully (already numerous ones had bitrotted, or
never worked in the first place due to typos).

The next step is to detect comments like "run with 'foobar' this will output
'barfoo'" and thus treat them like any other test.

Infrastructure:
- ccanlint is a little more verbose: -vv to show more, '-vvv' option to
  show *everything*.
- tests are now run in a temporary directory, so they don't need to use
  tmpfile()
- valgrind test failures are now reported!  (oops!)
- _info can specify tests which shouldn't be run (eg. not valgrind
  clean, or don't care about whitespace)
- ccanlint now tests that Example: code fragments compile!

New modules:
- compiler - one place for compiler-specifics (eg. __attribute__).
- opt - simple command line parsing 
- tally - running tally of integers 

Existing modules:
- tdb2 - Serious amounts of work and a rewrite, it is basically feature complete
         bar transactions.
- tdb - updates, can use jenkins hash now (in sync with SAMBA version).
        Now uses ccan/tally for tdb_summary (with graphs!)

Junkcode:
- New endianness test program (but see config.h's HAVE_BIG_ENDIAN)

Cheers!
Rusty.


More information about the ccan mailing list