stephan
|
009601d534
Remove $prefix/include from the default -I path because it can cause the build to pick up an unintended copy of sqlite3.h. Extend the ICU configure support (the origin of -I$prefix/include) to enable fetching the -I path from icu-config and apply it only to those objects which need it.
|
1 month ago |
stephan
|
c096d6add6
Add --enable-dev configure flag which sets various other flags.
|
1 month ago |
stephan
|
f154cef8f2
Document the if block at the end of sqlite-check-tcl.
|
1 month ago |
stephan
|
3b56698541
configure: avoid performing multiple checks for -lm on behalf of --enable-fts4 and --enable-fts5.
|
1 month ago |
stephan
|
45c2b1804f
Rename proj-define-if-opt-truthy to the more accurate, and less verbose, proj-define-for-opt.
|
1 month ago |
stephan
|
e9b04e524b
Simplify usage of proj-assert by making the expr prefix implicit. Add an optional description arg to proj-assert, defaulting to the body of the assertion script.
|
1 month ago |
stephan
|
497f84885b
Remove an unused lib import from auto.def.
|
1 month ago |
stephan
|
a700db9387
Add 'tcl' makefile target which builds all but tclextension (which does not have a well-defined name and dependencies). Improve the deps for .tclenv.sh to avoid getting a stale tclsh when re-configuring with a different --with-tcl(sh).
|
1 month ago |
stephan
|
6f95d95015
Disable setting of the SONAME (enabled by [2a2419ef742]), as it's not clear whether blindly setting the SONAME, which now differs from its historical value, will cause more grief than it solves. Add a (disabled) experiment which permits setting (or not) the SONAME to the legacy or current values. This change is up for further change as experimentation proves whether we truly need/want the SONAME. See discussion in/around [forum:0c6fc6f46b2cb3|forum post 0c6fc6f46b2cb3].
|
1 month ago |
stephan
|
ee9950d09b
Extend the set of --flags which get cleared/unset when using --with-wasi-sdk.
|
2 months ago |
stephan
|
6e6820148a
Doc touchups in the line-editing feature check code.
|
2 months ago |
stephan
|
c7882b975e
tcl configuration: --with-tcl=prefix is equivalent to passing the --prefix dir to it. If --with-tcl or --enable-tcl are explicitly passed in and tclConfig.sh is not found, fail fatally. When TCL is either explicitly disabled or default search for it fails non-fatally, be more explicit about which components are not available.
|
2 months ago |
stephan
|
4c3139d7ee
Add some internal docs to auto.def explaining certain passages.
|
2 months ago |
stephan
|
00e863c48e
Document the purpose of a global auto.def var added in [0d558318172d].
|
2 months ago |
stephan
|
d9345f3312
When --with-linenoise refers to the msteveb flavor and jimsh is the TCL used for code generation, enable linenoise in jimsh. Remove some dead auto.def code.
|
2 months ago |
stephan
|
968dc2489a
Change several 'file exists' checks for tclsh to file-isexec because that checks for .exe extensions on Windows.
|
2 months ago |
stephan
|
347e4833f9
For platforms where tclsh is found but tclConfig.sh is not, set HAVE_TCL to 0. We can't build the TCL components on those. Problem reported in [forum:5106aee3a8|forum post 5106aee3a8].
|
2 months ago |
stephan
|
6da414ef41
When constructing the auto-reconfigure commands, only quote args which look like they need it. Minor doc updates.
|
2 months ago |
stephan
|
4c2f7e57a2
Add autosetup/README.md - maintenance-related docs for SQLite developers (e.g. how to update autosetup). Start marking up the proj.tcl APIs with autosetup's doc markup so that they appear in the ./configure --reference output.
|
2 months ago |
stephan
|
e957bd1ce7
Enhance the --with-emsdk flag to use a default value of 'auto', meaning to search the environment for it, and to fail fatally if --with-emsdk is explicitly provided but the SDK is not found.
|
2 months ago |
stephan
|
d2ac1c5b7c
Remove the --enable-rbu flag from the configure script, per /chat discussion, because it's a highly niche feature not in active use.
|
2 months ago |
stephan
|
6d443b0d74
auto.def: remove an extraneous global var, some dead tcl/makefile code/comments, and some extraneous output. Rename an internal-use proc. Disable the ext/wasm cleanup in the top-level (dist)clean because the noise from gmake is irritating.
|
2 months ago |
stephan
|
e69d98df22
Enhance the rpath configure check to honor --exec-prefix=... Remove some debugging code and move other debugging code around.
|
2 months ago |
stephan
|
63346a4bca
Add the --enable-rbu configure flag and building of the commented-out rbu binary.
|
2 months ago |
stephan
|
2fd38836dc
Fix typo of --libexec ==> --libexecdir, discovered via audit of a downstream build script. Unrelated doc touchups.
|
2 months ago |
stephan
|
f3d279595e
When installing the shared lib, re-activate the accommodation of legacy-style .so links (removed in [80584e165e4]) based on two reports that not having those will cause all clients linked against the legacy naming convention to fail to dynamically link (which would preclude an update of a system-level libsqlite3 package). Set up the infrastructure needed for adding an SONAME to the library but do not yet activate it. See discussion in/around [forum:046133a7da|forum post 046133a7da].
|
2 months ago |
stephan
|
88349483ef
Move the .default-CFLAGS handling into a utility function, rename the corresponding file to .env-VAR, and apply that lookup to a couple more places.
|
2 months ago |
stephan
|
a9b00b0fb7
Fix a CFLAGS.jimsh typo in Makefile.in. Change default jimsh flags to include -O1, which slows down its build but speeds up its runtime considerably.
|
2 months ago |
stephan
|
97c709b5be
If any autotools-conventional dir names are explicitly overridden via ./configure --dirname=X then export those names as-is to the makefile, otherwise derive them from $prefix. This is, at least in part, a compromise to provide package maintainers a way to specify a libdir which will be incorporated into the -rpath link flag (which we can only determine at configure-time).
|
2 months ago |
stephan
|
f0e7f4c1f2
An experiment in hybridizing overriding of autotools-conventional XYZdir vars, the goal being to be able to provide both overrriding of them at configure-time and make-time with sensible semantics. Based on notes from [forum:00d12a41f7|forum post 00d12a41f7].
|
2 months ago |