ylib, branch master A collection of small, reusable C libraries http://g.blicky.net/ylib.git/atom 2015-12-15T01:47:21Z Add linked list convenience macros 2015-12-15T01:47:21Z Yorhel git@yorhel.nl 2015-12-15T01:47:21Z urn:sha1:185fa04e0418ea3831fc1d8fb1b20147a3993dd8 Also taken from Globster, with some additions. vec: Rename include-once macros to something more sane 2015-12-14T08:40:12Z Yorhel git@yorhel.nl 2015-12-14T08:39:28Z urn:sha1:b3167d9b1d2b0fefed2853ab1ab39ee9e146e727 The UTIL_VEC_H came from its use in Globster. Added vec.h + tests 2015-12-14T08:24:53Z Yorhel git@yorhel.nl 2015-12-14T08:24:53Z urn:sha1:d1dbb446000376edeec923ef611ddbb0f75c5040 Based on Globster's util/vec.h, but with some additions. evtp: Don't inherit signal mask in worker threads 2014-04-03T06:50:29Z Yorhel git@yorhel.nl 2014-04-03T06:50:29Z urn:sha1:77ecca991f6775668f7b4213fde7a7beb01575f4 Thanks to Marc Lehmann for pointing out this problem and suggesting the fix: http://lists.schmorp.de/pipermail/libev/2014q2/002368.html Rewrote evtp: + _maxthreads(), - _die(), no more killing idle threads 2013-06-29T10:37:47Z Yorhel git@yorhel.nl 2013-06-29T10:28:55Z urn:sha1:cdeba6e7e187d6295b319337067e8c445a662a3f The previous code would behave badly with respect to killing idle threads, so I removed that feature for now. Instead, I added evtp_maxthreads() to dynamically control the maximum number of threads. The main motivation for this function is to dynamically control the parallelism when hashing files in Globster, but it's also useful for (manually) killing idle threads and for pausing any progress for a while. I also simplified destruction into a single function: evtp_destroy(). This is a bit less confusing, and the old non-blocking evtp_die() wasn't too useful anyway because you don't get a notification on when the threads are actually dead. Similar non-blocking behaviour can be approximated with the current functions by using evtp_maxthreads(tp, 0), waiting for a while, and then calling evtp_destroy(). That, however, doesn't guarantee non-blockyness. ylog: Add <default_level> pattern as alias for a wildcard match 2013-06-04T14:48:23Z Yorhel git@yorhel.nl 2013-06-04T14:48:23Z urn:sha1:5a631446b8e8bf1a1dac5574aeafe9e0f36452b7 This simplifies the common case where you just want to specify a single log level for all files. Instead of writing "*:5", a simple "5" will do with this addition. ylog: Add tests for some internal functions + fix config parsing bug 2013-06-04T14:32:08Z Yorhel git@yorhel.nl 2013-06-04T14:31:29Z urn:sha1:a510c7269d2223110614047e1deed55f06c1b20d ylog: Fix off-by-one in ylog_enabled() + add ytrace() 2013-06-03T15:26:03Z Yorhel git@yorhel.nl 2013-06-03T15:26:03Z urn:sha1:f04b0a293d2f74ec454a331737738a0678f00f40 yuri: Set missing fields to empty string instead of NULL 2013-05-26T14:56:16Z Yorhel git@yorhel.nl 2013-05-26T14:56:16Z urn:sha1:c9bbce9c94a67ef60ed81c14ac82a69f89220c27 I can't see how the distinction between "empty" and "absent" is important. I've also updated the test suite to test that all pointers in the struct are set to point within the given buffer. This is important even for the empty string, because the pointed memory must be writable. Calling yuri_unescape() on an empty string allocated somewhere in read-only memory would cause issues. yuri: Rewrite query string parser to be smaller and more flexible 2013-05-26T09:29:23Z Yorhel git@yorhel.nl 2013-05-26T09:23:02Z urn:sha1:8aa2e407c7b357248c9a7cbaf124f26f5d327170 This is a much better API. You can only iterate once over a single string, but multiple iterations with the previous API wasn't going to be fast either way, so might as well enfore a string copy if you want that. This new API allows more not-so-very-well-formed query strings, too. Such as multiple ";&;&" sequences and empty/absent keys/values. Usually the goal is merely to extract information out of a URI, in which case such oddities can be safely ignored. In case the application wishes to disallow or validate such URIs, the new API can still be used for that but requires a bit more work.