notmuch/performance-test/perf-test-lib.sh, branch 0.17 thread-based email index, search, and tagging https://git.notmuchmail.org/git/notmuch/atom?h=0.17 2012-12-31T01:12:11Z perf-test: initial support for talloc leak report in memory tests 2012-12-31T01:12:11Z David Bremner bremner@debian.org 2012-12-26T16:29:45Z urn:sha1:8d05dcfffe001dc64d750f509e049340320b4dd9 As with the valgrind logs, we print a (very) brief summary and leave the log for inspection. perf-test: initial version of memory test infrastructure. 2012-12-25T12:49:24Z David Bremner bremner@debian.org 2012-12-16T12:33:17Z urn:sha1:098ef4af4d0a52a6b4daed5324a7c77f6c9108da The idea is run some code under valgrind --leak-check=full and report a summary, leaving the user to peruse the log file if they want. We go to some lengths to preserve the log files from accidental overwriting; the full corpus takes about 3 hours to run under valgrind on my machine. The naming of the log directories may be slightly controversial; in the unlikely event of two runs in less than a second, the log will be overwritten. A previous version with mktemp+timestamp was dismissed as overkill; just mktemp alone does not sort nicely. One new test is included, to check notmuch new for memory leaks. perf-test: unpack tags. 2012-12-15T12:17:58Z David Bremner bremner@debian.org 2012-12-04T02:59:56Z urn:sha1:5c35791d7cf7279024befc98e9d973af461336a3 There is only one set of tags, independant of the size of message corpus chosen. perf-test: add caching of xapian database 2012-12-15T12:17:58Z David Bremner bremner@debian.org 2012-12-03T12:48:53Z urn:sha1:ceaf5ca6c02e494eee5b9b9aa955459b3cd29471 The caching and uncaching seem to be necessarily manual, as timing the initial notmuch new is one of our goals with this suite. perf-test: cache unpacked corpus 2012-12-15T12:17:57Z David Bremner bremner@debian.org 2012-12-03T12:13:31Z urn:sha1:74a883562b3e4593c75fa7625ff5cabab46a6466 Unpacking is not really the expensive step (compared to the initial notmuch new), but this is a pre-requisite to caching the database. perf-test: optionally print description for each group of tests 2012-12-15T12:17:57Z David Bremner bremner@debian.org 2012-12-06T01:41:44Z urn:sha1:925ebd1fde59b003a7215e8cd7e07225a8f8915d Output from tests is indented slightly in the same style as the correctness tests. perf-test: add corpus size to output, compact I/O stats 2012-12-15T02:23:55Z David Bremner bremner@debian.org 2012-12-04T02:51:14Z urn:sha1:a3137c61e144935fde9896855ffc8625f7505c37 Austin suggested a while ago that the corpus size be printed in the header. In the end it seems the corpus will be fixed per test script, so this suggestion indeed makes sense. The tabbing was wrapping on my usual 80 column terminal, so I joined the input and output columns together. perf-test: add argument parsing for performance tests 2012-12-15T02:23:55Z David Bremner bremner@debian.org 2012-12-03T23:56:20Z urn:sha1:e7c661d31a23d1f711a84e6eca85eea9b67f11fe This patch just sets (non-exported) variables. The variable $debug is already used, and $corpus_size will be used in following commits. perf-test: propagate non-zero returns from /usr/bin/time 2012-12-15T02:23:54Z David Bremner bremner@debian.org 2012-12-06T01:37:29Z urn:sha1:e52597016f5e5dfbceb73068dd165f3f3849ee7b Unlike in the correctness tests, the most common cause of non-zero return seems to be the user interrupting, so killing the run seems like the friendly thing to do. performance-test: add units to In and Out 2012-11-29T13:28:46Z David Bremner bremner@debian.org 2012-11-28T21:27:01Z urn:sha1:a4eabfd1b7b6c72af09153df082bac3dca7c4420 Austin Read enough Fine Manuals to figure out this is in multiples of 512 bytes.