10 * Rather than looking up the linker's hardcode characteristics in a
11 table of shell code, use objdump or equivalent to probe a test program
14 * Eliminate the warnings from autoconf -Wobsolete.
16 * Hook the various language dependencies into the autoconf _AC_LANG
19 * Work out what to do when the dynamic linker loads needed dependencies.
21 * We could have an option to hardcode paths into libraries, as well as
22 binaries: '... -Wl,-soname -Wl,/tmp/libtest.so.0 ...'. This is not
23 possible on all platforms, and is in part obviated by the ability of
24 linking libtool libraries specified with -lname, but it might still
27 * Lists of exported symbols should be stored in the pseudo library
28 so that the size of lt_preloaded_symbols can be reduced.
30 * Have some option to tell libtool not to include -L flags that point
31 into a certain tree in the dependence list of an installed library.
32 For example: -L-$top_builddir would let one link with libtool
33 libraries in sibling subdirectories within a project, using the -L
34 notation, without getting builddir pathnames ever mentioned in .la
35 files that get installed.
37 * Eric Lemings <elemings@cyberia.lemings.com> writes:
38 Because of a growing number of config scripts for packages in GNOME 1.2
39 (e.g. glib-config, xml-config, orbit-config. etc), development of GNOME
40 2.0 spawned a separate tool called pkg-config that allows all packages
41 to use one tool rather than several different scripts to query compile
42 flags, link flags, and other configuration data.
44 The functionality of pkg-config seems to me to have a lot of overlap
45 with the goals of libtool. I was wondering if anyone had considered
46 adding an eighth mode to libtool that just queries the installed
47 library for the same information that pkg-config provides. Since
48 most packages that use pkg-config also use libtool, I think this
49 would be a good way to reduce maintainer and developer dependencies.
51 * Have libtoolize install 'install-sh' if a newer version is available,
52 and/or Automake is not used.
54 * Allow to specify linking some dependent libraries statically and some
55 dynamically, where possible.
57 * Improve support for C++ with templates.
59 * Audit file listing in libtool.m4.
61 * Fix deplibs_check_method=pass_all (which is wrong!) on GNU/Linux.
63 * Fix -dlopen "self" on AIX. Reported by Gary Kumfert <kumfert@llnl.gov>.
65 * Fix denial of service if using installed 'libtool' on a different mount point
66 together with a compiler that does not understand '-c -o'.
67 Reported by Marcin Siennicki.
69 * Look at better -no-undefined support, maybe along the idea of
70 [support #103719] for CC.
76 * Change libltdl interface: add separate functions for function
77 pointers. This will allow porting to systems where function pointers
78 are incompatible with data pointer C-wise.
80 * Fix the following bugs in libltdl:
81 - error reporting of tryall_dlopen():
82 if the file actually doesn't exist (stat() fails or it wasn't dlpreopened)
83 -> report 'file not found'
84 if it cannot be loaded (e.g. due to missing dependencies)
86 open question: what error should be reported if all dlloaders fail
87 or if a specific module type can only be loaded by one of them, how report its dlerror?
88 Also report dlerror() for dlclose and dlsym if available
89 - Make sure that the dependency_libs of a dlpreopened module won't be loaded.
91 - Fix mdemo failures on mingw.
93 - Fix the last memleak. Reported by Jeff Squyres <jsquyres@lam-mpi.org>.
95 - Fix LTDL_CONVENIENCE. Reported by Bob Friesenhahn
96 and Patrick Welche <prlw1@newn.cam.ac.uk>.
102 * Rewrite the func_copy_* functions so that instead of forking 2 tar
103 processes per copied file, a list of files to copy is built and all
104 files copied with a single pair of tar processes.
106 * Write test case that adds libtool macros to aclocal.m4.
115 * Need to finalize the documentation, and give a specification of
116 '.la' files so that people can depend on their format. This would be
117 a good thing to put before the maintainance notes.
119 * Document the installed 'libtool' and its limitations clearly (maybe implement
120 --disable-script-install as well). Or, even better, remove its limitations.
122 * Platform notes redo.
127 * Rewrite the whole thing in Autotest. This will enable us to remove
128 all the tests/*demo noise, and duplication; and thus speed up bootstrap
129 and make writing new tests a whole lot more pleasant.
131 * We should include tests with reloadable objects in the testsuite.
133 * Write a test case for linkage with gnu ld scripts (per 2004-08-25 patch
139 - multiple input files
144 * Fix cross-compiling.
148 * Support multilibbing.
150 * If not cross-compiling, have the static flag test run the resulting
151 binary to make sure everything works.
153 * Another form of convenience library is to have undocumented utility
154 libraries, where only the shared version is installed.
156 * We could use libtool object convenience libraries that resolve
157 symbols to be included in a libtool archive. This would require some
158 sort of -whole-archive option, as well.
160 * Currently, convenience libraries (.al) are built from .lo objects,
161 except when --disable-shared. When we can build both shared and
162 static libraries, we should probably create a .al out of .lo objects
163 and also a .a out of .o objects. The .al would only be used to create
164 shared libraries, whereas the .a would be used for creating static
165 libraries and programs. We could also explicitly support 'empty'
166 convenience libraries, that behave as macros that expand to a set of
167 -Rs, -Ls and -ls switches.
169 * Audit use of object names so we can allow '$' not only within
170 source file names. Necessary especially for java.
172 * We could introduce a mechanism to allow for soname rewriting, to
173 ease multi-libc support. Installers could specify a prefix, suffix or
174 sed command to modify the soname, and libtool would create the
175 corresponding link. This would allow for rebuilding a library with
176 the same version number, but depending on different versions of libc,
177 for example. In the future, we might even have an option to encode
178 the sonames of all dependencies of a library into its soname.
180 * Look again at a binary C libtool, or byte-compiled libtool to improve
183 * Generate some "platform specific" shell functions with config.status,
184 for example, there is no need to have the C source code for the
185 wrapper script on non-windows platforms, this will make the generated
186 libtool script smaller and easier to follow, maybe a little faster
189 * Audit the GCJ tag section in libtool.m4.
191 * Add caching mechanism. Look at 'libtool-cache' from Robert Ă–gren.
194 2.4. libtool autoconf macros
195 ----------------------------
197 * Sort out the macro mess in libtool.m4. We've started this already
198 by refactoring chunks into separate files, but I never did completely
199 untangle the mess of macros imported from ltconfig.
201 * The definitions for LT_SYS_MODULE_EXT, LT_SYS_MODULE_PATH and
202 LT_SYS_DLSEARCH_PATH should not rely on the _LT_SYS_DYNAMIC_LINKER
203 macro. This involves moving the code that sets the variables
204 library_names_spec, shlibpath_var and sys_lib_dlsearch_path_spec from
205 into a separate macro, and AC_REQUIRING the newly extracted macro in the
206 respective ltdl.m4 macros.
208 2.5. libtool automake integration
209 ---------------------------------
211 * Unify locks between libtool and compile.
218 * Finish the rewrite of the core libltdl. The loaders are fine, and
219 the outlying code is now good. Ralf is starting to pick away at a lot
220 of the remaining nasties already, but the code for finding .la/.so files
221 and reading/loading them could use a lot more improvement.
223 * I think we could factor out a little path management support module
224 from existing libltdl. This would be useful for M4 at least -- keeping
225 track of FOO_PATH environment contents, searching for files in paths
228 * Try to find a work-around for -[all-]static and libltdl on platforms
229 that will fail to find dlopening functions in this case. Maybe
230 creating an alternate libltdl that provides only for dlpreopening, or
231 creating an additional static library to provide dummy implementations
232 of the functions that can't be linked statically. This could hardly
233 be made completely transparent, though.
235 * In conjunction with above, fix the failures on *BSD when linked to
236 static libc. Reported by Guilhem Lavaux <guilhem@kaffe.org>.
238 * Add i18n strings to libltdl, ensuring that package developers can
239 ignore any i18n when they libtoolize.
244 * Arrange that EXEEXT suffixes are stripped from wrapper script names
245 only when needed, and that a timestamp file or a wrapper program is
246 created with the EXEEXT suffix, so that 'make' doesn't build it every
249 * Figure out how to use data items in dlls with win32.
250 The difficult part is compiling each object that will be linked with an
251 import lib differently than if it will be linked with a static lib. This
252 will almost definitely require that automake pass some hints about linkage
253 in to each object compilation line.
255 * jeffdb@goodnet.com writes:
256 all you need to do for mutually dependent .dll's is to create an implib from
257 a .def file so it appears that we might need to detect and handle mutual
258 dependencies specially on win32 =(O|
260 * QoI for file name and path conversion functions. Currently, these are
261 implemented as MxN different functions; this has quadratic complexity. If
262 possible, it would be preferred to implement then as M+N functions. However:
263 http://lists.gnu.org/archive/html/libtool-patches/2010-08/msg00224.html
264 The main issue is you don't know what the "native" (e.g. "central")
265 path-type is; e.g. "from-X (to what?)" and "(from what?) to-Y". Right
266 now there are only four "platforms" involved: *nix, mingw, msys, and
267 cygwin. That's actually the break-even point, given the vagaries and
268 optimizations involved in these particular four platforms.
270 We have exactly five basic file name conversion functions (not counting
271 the wrappers that handle paths). For a non-quadratic M+N (from-X|to-Y)
272 solution, we'd need, I think, the same number of conversion functions: brute
273 force suggests nine (four to_*, four from_*, plus the noop), but then many
274 of the from_* and to_* would actually BE noop.
276 I'm assuming here that the "central" path-type is implicitly some sort of
277 unixish -- maybe cyg, maybe msys, maybe unix -- path-type. The issue is
278 that each of the five conversion functions use a different TOOL to perform
279 the conversion, with different syntax. So, trying to combine, e.g.
283 into an all-encompassing "central_unixish_to_mingw" would require additional
284 m4 magic to basically replace the guts depending on whether $build was msys,
285 cygwin, or unix. Worse, you can't really do a set of
286 {msys|cygwin|unix}_to_central_unixish that isn't simply a no-op -- because
287 (A) they already are all unixish, and (B) what tool would you use? How would
288 the later to_mingw function "know" how to covert this new representation to
289 mingw. So, {msys|cygwin|unix}_to_central_unixish would simply be a no-op
290 and central_unixish_to_mingw would still do all the work (with its guts
291 customized based on $build).
293 For more reasonable cross environments (e.g. linux-gnu->some_embedded) I
294 think you could probably work out a general M+N scheme, since most embedded
295 $hosts aren't as strange as the win32 variants -- even VxWorks and INTEGRITY
296 have basic, unix-like file systems (although INTEGRITY does have multiple
297 roots). Aggressive use of the m4 function_replace machinery WOULD be
298 appropriate for /these/ conversion functions. OTOH...(a) you can't run the
299 $host apps on $build anyway, in these embedded situations. At best you'd use
300 $TARGETSHELL and "run" them via a remote connection, and (b) they don't use
303 So...I don't think it makes much difference *right now* in the amount of
304 code required, or the number of functions implemented. At some point in the
305 future we might want to generalize to an M+N scheme. For the existing win32
306 $hosts, all of the funtionality would be on "one side" of the 2-step
307 conversion; the "other side" would be noop. But we won't worry about the
308 implicit quadratic complexity of the existing scheme for now.
313 * Maybe implement full support for other orthogonal library types
314 (libhello_g, libhello_p, 64 vs 32-bit ABI's, etc). Make these types
317 * Perhaps the use of libltdl could be made cleaner by allowing
318 registration of hook functions to call at various points. This would
319 hopefully free the user from having to maintain a parallel module
320 list with user data. This would likely involve being able to carry
321 additional per user module data in the lt_dlmodule structure -- perhaps
322 in the form of an associative array keyed by user name?
324 * Figure out how to make pkg-config aware of the information libtool
325 knows about libraries and their dependencies, and send a patch.
327 * Generate a libtool.m4 from a bunch of individual files, one per
328 platform, to make the job of a "platform maintainer" easier and make
329 it easier to add new platforms.
332 Copyright (C) 2004-2005, 2007-2008, 2011-2019, 2021-2024 Free Software
334 Written by Gary V. Vaughan, 2004
336 This file is part of GNU Libtool.
338 Copying and distribution of this file, with or without modification,
339 are permitted in any medium without royalty provided the copyright
340 notice and this notice are preserved. This file is offered as-is,
341 without warranty of any kind.