libcwal  Timeline

Many hyperlinks are disabled.
Use anonymous login to enable hyperlinks.

50 check-ins occurring around 45c756c35197991b.

2014-05-09
15:09
a valgrind cleanup fix. check-in: [b489e11b5b] user: stephan tags: trunk
15:09
reenabled an old assertion. check-in: [05885cd8d9] user: stephan tags: trunk
15:07
added th1ish_autosweep_config() and tweaked the shell a bit to use it. check-in: [a9ae99514e] user: stephan tags: trunk
15:04
Test code tweaks. check-in: [f7b8731bd6] user: stephan tags: trunk
15:04
minor internal tinkering. check-in: [98dee75be2] user: stephan tags: trunk
12:12
test code improvements. check-in: [bbeff509a7] user: stephan tags: trunk
11:31
Found the hashing (int range) culprit which broke interning lookups. Other tweaking with vacuuming and interning. Added a kludge ++refcount on interned strings upon insertion to save them from vacuuming. i don't like that, but it does save memory. Switched string hash algos again. check-in: [a6668115e9] user: stephan tags: trunk
10:38
More work on vacuuming and related code. Discovered in the process that string interning is not hitting anywhere nearly as often as i really expect it to. check-in: [a24cf62be2] user: stephan tags: trunk
2014-05-08
23:41
Refactored cwal_engine_vacuum() to... drum roll, please... no longer copy anything. It now only rescopes stuff, so no allocation needed, which means no error cases unless it detects memory corruption (assertion). check-in: [ae032f2d59] user: stephan tags: trunk
23:13
minor cleanups before larger changes. check-in: [65535570a0] user: stephan tags: trunk
23:04
Started adding (untested, except that existing code/tests still work) infrastructure for making individual values, and those they reference, safe from vacuuming up. Intention: a way to make non-script-visible values safe from vacuum. check-in: [f5ae60cad0] user: stephan tags: trunk
2014-05-07
22:03
tinkering with save/restore of "this", working towards fixing the bug revealed in the interactive linenoiseish. Bed time. check-in: [100efdd360] user: stephan tags: trunk
21:11
improvements in the ++/-- handling. Still an ie->currentThis assertion triggering in linenoiseish interactive mode which does not trigger in "normal" scripts. check-in: [371ec25cc6] user: stephan tags: trunk
21:10
minor cleanups. check-in: [c98c402d07] user: stephan tags: trunk
19:29
metrics output is now less cramped. check-in: [2a5a4c7ff8] user: stephan tags: trunk
19:03
Now stashes more files after each valgrind run. check-in: [d78e840f5c] user: stephan tags: trunk
19:02
Automated unit tests now run scripts in verbose mode and capture their output to a file. Clean up ms_print temp files. check-in: [aa28f83f4e] user: stephan tags: trunk
19:01
cwal_dump_allocation_metrics() now counts scopes differently (only allocated ones) because they skew the stats unduly (and unfairly ;). check-in: [e4e693eb9f] user: stephan tags: trunk
17:38
typo fix, upped the version count to 4. check-in: [f955b94eca] user: stephan tags: trunk
17:15
added vgLastFewVersions.sh. check-in: [0879b36247] user: stephan tags: trunk
2014-05-06
20:59
Optimized auto-sweep to replace every N (configurable) sweeps with a vacuum instead. Vacuum doesn't actually cost as much time as i thought: adds about 5ms runtime when all current unit tests are piped through a single interpreter instance. Aggressive (100%) vacuuming saves only a tick of memory but it is easy to envision cases for which it is more... check-in: [94843b2b50] user: stephan tags: trunk
20:54
minor cleanups. removed some dead code. added some comments. check-in: [571cccbd37] user: stephan tags: trunk
2014-05-05
22:24
moved api namespace setup to after flags setup - the string "api" is still not interning for reasons i do not understand. Minor defensive th1ish_read_identifier() tweak. check-in: [d6a0881e64] user: stephan tags: trunk
22:23
minor tracing message tweak. check-in: [f97d4ed9e2] user: stephan tags: trunk
19:54
minor test code tinkering. check-in: [b300a53e7a] user: stephan tags: trunk
19:38
Vacuum now works as a drop-in replacement for sweep in th1ish :-D meaning that, in principle, it is (with proper script structuring) possible to create a leak-free, really-long-running scope, regardless of cycles. :-D! Interestingly, vacuum does not often catch more values than sweep, so it will be optimized to normally sweep, and replace every Nth... check-in: [4ce80bab79] user: stephan tags: trunk
18:06
this variant of cwal_engine_vacuum() works for the first 3 th1ish unit test scripts, then falls apart somewhere, but is certainly better than the previous iteration. check-in: [7a0a0b3e7a] user: stephan tags: trunk
2014-05-04
15:43
if/else blocks now resolve to true if any IF part was entered, else false. Improvements in tmplish. check-in: [a162cdb1b4] user: stephan tags: trunk
12:46
more work on the templates. Fixed a scope/eval lifetime bug which just now surfaced for some reason. check-in: [c2b6cda836] user: stephan tags: trunk
12:36
minor internal cleanups. check-in: [4123cb68de] user: stephan tags: trunk
11:07
Started work on support for creating scripts from text docs containing markup tags, similar to PHP/th1. check-in: [3d88a70d24] user: stephan tags: trunk
11:07
Added long-overdue cwal_buffer_clear() convenience function check-in: [4dd397b5b4] user: stephan tags: trunk
01:57
Added a flag mechanism to change HEREDOC parse behaviour a bit: <<<:X ... X only strips a single space/newline from the beginning/end. Simplifies some formatting in template processing and fossi1ish. check-in: [0d8f9346fa] user: stephan tags: trunk
00:07
Fixed a comma parsing bug when $call syntax is used as a value in an object literal. check-in: [b93e7a08e4] user: stephan tags: trunk
2014-05-03
23:51
more tinkering with cwal_engine_vacuum(). check-in: [0cad6273e0] user: stephan tags: trunk
20:24
Fixed test to not result in an exception (last value eval'd) as it looks like an error in the output. check-in: [1ec7bd5b29] user: stephan tags: trunk
20:08
Added postfix ++/-- ops. Only lightly tested. check-in: [a55c52113f] user: stephan tags: trunk
20:07
Added cwal_prop_flags_v(), admittedly a quick-hack measure for th1ish which needs to be improved upon at some point. check-in: [73df633530] user: stephan tags: trunk
19:35
Added cwal_prop_take_v(), though it has kinda funny semantics for the key part. Was necessary for postfix obj.prop++/-- operators. check-in: [01628d5381] user: stephan tags: trunk
18:21
added unit tests for prefix ++/-- ops. check-in: [9971f84d4a] user: stephan tags: trunk
18:16
Implemented PREFIX ++/-- for identifiers and object.property access. Not yet sure it's relatively bug-free, so it's not yet documented. check-in: [93fa84de08] user: stephan tags: trunk
18:10
removed reference to a removed function. check-in: [29aed27ef1] user: stephan tags: trunk
16:28
More tinkering with vacuum. It does horrible things to th1ish. check-in: [93310a20c0] user: stephan tags: trunk
15:36
added function call()()() chaining test. check-in: [ff016188af] user: stephan tags: trunk
15:35
more work on vacuum. Seems it cannot be dropped right in to th1ish for reasons as yet determined but very possibly incompatibilities with vacuuming assumptions. check-in: [e276ed0f1d] user: stephan tags: trunk
14:59
doc improvements. check-in: [8079a2c009] user: stephan tags: trunk
14:53
forgot to check in first cwal_engine_vacuum() test code. check-in: [7d7ad9bb3a] user: stephan tags: trunk
14:33
Introducing cwal_engine_vacuum(), an experimental algo which cleans up all non-variable values which are themselves not reachable via variable refs/containers. Fixed two bogus assertions while doing so. check-in: [8f455e260b] user: stephan tags: trunk
11:45
Fixed chain(function)(calls). Might have side-effects but the unit tests (and those in libfossil) still run. check-in: [2093a87333] user: stephan tags: trunk
10:20
enabled unit tests for +/-DOUBLE_STRING conversion. check-in: [c4db2e502e] user: stephan tags: trunk