Many hyperlinks are disabled.
Use anonymous login
to enable hyperlinks.
50 events by user dave occurring around 2014-10-20 17:09:33.
|
2014-10-25
| ||
| 19:52 | correct silly position of assert(); correct a comment; correct for sqlite3 not auto-initing around mutex code (for reasons unknown to me) check-in: 455133309f user: dave tags: dave | |
| 18:46 | previous idempotency with atomic inc/dec isn't really enough; the action being protected needs to be guarded as well. plus there were link issues on linux, the resolution of which is machine architecture specific, so I ditched that approach. check-in: 1be04c571a user: dave tags: dave | |
| 18:17 | add atomic functions around refcounted init of sqlite extensions; unregister extensions when last one is freed check-in: 31a1e741ae user: dave tags: dave | |
| 16:28 | merge trunk, get stephans various fixes check-in: 84b7b9a10a user: dave tags: dave | |
|
2014-10-24
| ||
| 13:25 | prevent changing of sepstyle for non-multivar types on update; fix constraint check on insert check-in: e7011739c4 user: dave tags: dave | |
| 05:12 | when inserting versioned settings, check metadata 'is multivar' and force attribute in new settings object accordingly check-in: 9c9c3b6069 user: dave tags: dave | |
| 03:20 | expose 'isversionable' settings metadata via vtable for those who wish externally to know. check-in: 2ac6a110d4 user: dave tags: dave | |
| 02:24 | integrate feech-versionedsettingsupport. Not strictly fully tested, but I feel pretty good about it, so I'm going to close it out. I'll do any bugfixes here that might show up in the next days of testing. check-in: 03c3149b0e user: dave tags: dave | |
| 01:22 | added support for 'non-multi-var settings' (most settings, actually), which inhibits list processing (i.e. doing comma/newline normalization). Also made alteration to VC6 project file to suppress warnings about sqlite3.c. Closed-Leaf check-in: 201339bcc3 user: dave tags: feech-versionedsettingsupport | |
|
2014-10-23
| ||
| 22:21 | bug storing match_or_fail context structure; modded some accept regexes to include the 'empty is OK' metametacharacter check-in: 699fec5042 user: dave tags: feech-versionedsettingsupport | |
| 21:29 | added more regex 'accept' patterns to settings metadata check-in: 82420a0765 user: dave tags: feech-versionedsettingsupport | |
| 19:57 | minor booboo with signedness; pleasure gcc check-in: 11afb2ec7a user: dave tags: feech-versionedsettingsupport | |
| 19:52 | support extended regex syntax for case-insensitive and multivalue settings check-in: 9660008e7b user: dave tags: feech-versionedsettingsupport | |
| 05:05 | can't use FSL_CKOUT_DIR() to coin the name of the versioned settings path, because sql evaluation is inhibited in the parameter list to a CREATE VIRTUAL TABLE, so I needed to coin the path manually. check-in: 7f14cc89a4 user: dave tags: feech-versionedsettingsupport | |
| 03:33 | fix a few dreaded // comments check-in: d4426ffe90 user: dave tags: feech-versionedsettingsupport | |
|
2014-10-22
| ||
| 21:48 | first attempt at wiring in virtual tables for settings metadata and versioned setting support. Also, added an aliased column 'value' to 'default_value' for the settings metadata, to allow existing query scheme to work consistently with that table as well. check-in: ddb939fd3f user: dave tags: feech-versionedsettingsupport | |
|
2014-10-21
| ||
| 19:50 | can build on gcc on Linux in conventional and amalgamation. now to test on other platforms... check-in: 6f02a467b1 user: dave tags: feech-versionedsettingsupport | |
| 01:51 | make static linkage on internal functions and data of settings vtable since we don't have anonymous namespaces in C. check-in: 2bcec54383 user: dave tags: feech-versionedsettingsupport | |
| 01:09 | more fun gcc warnings changes. I want my gcc. check-in: 136617e1c8 user: dave tags: feech-versionedsettingsupport | |
|
2014-10-20
| ||
| 23:07 | needed to 'namespace' the regex engine in regexp -- even parts declared 'static', because in amalgamation, there effectvely is no 'static'. check-in: be9ccb0711 user: dave tags: feech-versionedsettingsupport | |
| 18:03 | strdup is not ansi, so used fsl_strdup check-in: e26a27a507 user: dave tags: feech-versionedsettingsupport | |
| 17:49 | more tasty gcc warning accomodations check-in: 7e6a260c1a user: dave tags: feech-versionedsettingsupport | |
| 17:41 | a few more warning reductions for gcc check-in: e1aa9edc17 user: dave tags: feech-versionedsettingsupport | |
| 17:27 | missed a spot with some // to /**/ conversion check-in: d4733ab715 user: dave tags: feech-versionedsettingsupport | |
| 17:09 | sqlite3ext.h should not have been included unconditionally -- it is specific to dlls where host app provides sqlite impl. sqlite3ext.h needed to be added to repo, nonetheless, for some configurations. As such, sqlite3.h,.c were updated to be consistent with that file version (and also the final released version of 3.8.7) check-in: 0ab8c75c49 user: dave tags: feech-versionedsettingsupport | |
| 16:43 | added source to Makefile.in for non-msvc builds check-in: 1bc542834a user: dave tags: feech-versionedsettingsupport | |
| 13:06 | I committed into wrong branch, so pulling to the intended one. The changes thus far are benign, so I'm going to leave them there for the moment, and also since ultimately I'll be merging them back. (I'm such a buffoon.) check-in: d9afd076f1 user: dave tags: feech-versionedsettingsupport | |
| 13:02 | interim; add source for settings vtable and regexp extension functions; not wired it yet. update VC/DS projects. minor mod to auto.def to avoid redefining a define that oftentimes comes from the command line to cl, and would otherwise generate a warning in some projects. check-in: 4e32063e22 user: dave tags: dave | |
|
2014-10-19
| ||
| 15:31 | Create new branch named "feech-versionedsettingsupport" check-in: 5f632a80f1 user: dave tags: feech-versionedsettingsupport | |
| 15:07 | (keeping up to date; merge trunk) check-in: 92fd52048b user: dave tags: dave | |
|
2014-10-18
| ||
| 00:52 | pulled in current trunk prior to setting up for work on integrating some sqlite3 extensions (mainly for presenting an access method for versioned settings consistent with db-base settings, but also to unify settings metadata, and provide data validation for the settings values). I'll do a brief sanity check build and resolve anything that comes up there before making that feature branch. check-in: 5685461788 user: dave tags: dave | |
|
2014-10-02
| ||
| 03:15 | merged in dave's commits for validated msvc support. tested are vc6, ds2010, ds2012, dll, amalgamation, debug, release, x86, x64, and all possible permutations thereof. tested == all f-apps run to completion. check-in: c5218050d1 user: dave tags: trunk | |
| 03:12 | amalgamation support for msvc; #undef'ed a couple macros that collided with Windowsian things; added a project for my testing porpoises; put in a conditional compile switch in f-sanity.c to support an amalgamation build with ease. check-in: 0d2451c9c2 user: dave tags: dave | |
| 01:27 | added BUILDING document for the msvc builds. added DS2012 f-apps configs. check-in: 517c43c434 user: dave tags: dave | |
|
2014-10-01
| ||
| 15:36 | (pull in current trunk) check-in: ce9a57463c user: dave tags: dave | |
| 15:34 | integrated feature branch 'feech-dave-msvcsupport001' into my working branch; this mostly because it ws merged into trunk already, so I might as well close it. its was veritably done, pending final verification, which I expected to go cleanly anyway. check-in: f81b965c3a user: dave tags: dave | |
| 07:58 | minor; a warning is issued by gcc when it couldn't figure out if a variable might be used before it is initialized. It is not, so I redundantly initialized it to avoid the warning. Closed-Leaf check-in: 2a1b0091d2 user: dave tags: feech-dave-msvcsupport001 | |
| 07:36 | trivial; avoid a warning about there being a difference between void fxn() and void fxn ( void ) for old c compilers check-in: 6f99aa0874 user: dave tags: feech-dave-msvcsupport001 | |
| 07:31 | minor; avoid a warning arising from a conflicting definition of tzset check-in: 4a8d9c2548 user: dave tags: feech-dave-msvcsupport001 | |
| 07:27 | minor; the 'release' config of f-apps/test output to the wrong directory, with the wrong name check-in: f3c33050b9 user: dave tags: feech-dave-msvcsupport001 | |
| 07:20 | marked as FSL_EXPORT all other symbols in public api (pretty much everything). I stopped short of doing things in fossil-internal.h, because I think that is a private header. There are a couple things in there that are exported, because the test apps required them, but I'm going to leave that header alone for now. Changes in this change set reflect methods that are not referenced by any of the test apps, and maybe need some test cases generated. check-in: bd0a181022 user: dave tags: feech-dave-msvcsupport001 | |
| 06:15 | marked as FSL_EXPORT various symbols; this change set is limited to the ones that are needed by the various f-apps, but this is not complete for general use of the library. I am doing this change in two phases simply because it gives me a way to see what methods have been exercised (by f-apps) and which have not. check-in: 7b651689f3 user: dave tags: feech-dave-msvcsupport001 | |
| 06:10 | added project configs for the f-apps for VC6 and DeveloperStudio-2010 check-in: 62fc412621 user: dave tags: feech-dave-msvcsupport001 | |
| 05:59 | updated project config to use dll crt; this is required for some apis (e.g. some that transport FILE* created in the client app, but used in the lib, e.g. fsl_output_f_FILE()) check-in: b9562c9af7 user: dave tags: feech-dave-msvcsupport001 | |
|
2014-09-30
| ||
| 19:36 | added DevStudio 2012 configuration. This is pretty much the same as DS2010, but a minor change is made so as to produce binaries that will work on XP, and to turn off safe SEH for Win32 Release, since a dependent lib (zlib) can't produce safeSEH images (alas). check-in: c91f4e2770 user: dave tags: feech-dave-msvcsupport001 | |
| 15:35 | minor; workaround fact that VC6 doesn't know how to convert a uint64_t to a double, but it does know how to convert a int64_t to a double; haha. I didn't bother to put it under conditional compilation, since it is harmless on other platforms (and you can't get enough precision out of a double to ever have it matter about the highest bits). check-in: 4627c84c02 user: dave tags: feech-dave-msvcsupport001 | |
| 13:06 | corrected a bug in test_fs_cx_stat() that I introduced by commenting out an ureferenced local, simply to suppress a warning. It is unreferenced in Win32 builds, but it is used in Linux, so I uncommented it out, and I will just disregard the warning in Win32, since this is a test app anyway. Maybe I'll fix it better later with a conditional compile, or something, but it's pretty innocuous. check-in: 4814bd4e5c user: dave tags: feech-dave-msvcsupport001 | |
| 12:50 | a peculiarity of Microsoft's AMD-64 compiler is that __FILE__ resolves to a relative path correct at compile-time. but incorrect at runtime if you are running in a different directory. (The x86 compiler resolves to fully-qualified paths, which can create a different problem if you were on a different machine, but these are just test apps anyway). The net result was to cause tests to fail when __FILE__ was used as a object of the test. There were two test apps that did this, and they were modified to use a manhandled self name in the case of microsoft's compilers. check-in: 9834d60958 user: dave tags: feech-dave-msvcsupport001 | |
| 12:31 | stat() family of functions on Windows cannot have a trailing path separator(s); if they are present, ENOENT will be indicated. remove them before issuing the stat() call. check-in: 5f15aad639 user: dave tags: feech-dave-msvcsupport001 | |
| 12:22 | minor, win32dll; need to include header (fossil-util.h) declaring exported functions, or the impl will not actually get exported (the impl will not pick up the export attribute) check-in: ea7ba15dec user: dave tags: feech-dave-msvcsupport001 | |