| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
|
|
|
|
|
|
| |
Replace with a cross-reference to other documentation. We should not
duplicate documentation in the tests in general: the tests should cite
documentation (rather than code!) but the (necessary) duplication already
exists between implementation and documentation.
|
|
|
|
|
|
|
| |
Mostly to use C99-style declarations, moving declarations just before first
use.
Also make some code simplifications.
|
|
|
|
|
| |
Since the tests use their own isolated configuration directories, they no
longer risk polluting personal dictionaries.
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
list_dicts
Both dispose and dispose_dicts are required. This makes it harder to have
inadvertent leaks, and simplifies lib.c a little.
Require one of dictionary_exists or list_dicts. This means that request_dict
is no longer called by enchant_provider_dictonary_exists, so remove the
tests for this case.
|
| |
|
|
|
|
| |
Specifically enchant_{broker,dict}_get_error.
|
|
|
|
|
| |
Except for some deleted tests, now that it’s no longer possible to override
the system configuration directory.
|
|
|
|
|
|
|
|
|
| |
Assume a malloc-allocated char **.
Voikko used an internally-allocated char **, so copy it before passing it
back, as all the other providers have to.
In the process of fixing some tests, reactivate some disabled tests.
|
|
|
|
| |
for DLLs
|
| |
|
|
|
|
|
|
|
| |
Had to patch valgrind-tests.m4 to work around false positive in bash.
See https://lists.gnu.org/archive/html/bug-bash/2017-04/msg00038.html
Also add a suppressions file for relocatable.c.
|
|
|
|
|
| |
Some of these were intentional, but it seemed more sensible to change the
tests than either leave the leaks or write Valgrind suppressions for them.
|
| |
|
| |
|
|
|
|
| |
In order to remove Makefile.am duplication
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
| |
This adds a new API enchant_set_prefix_dir, which necessitated changing the
C++ API so that rather than constructing a broker instance statically, which
would pre-empt any setting of the prefix dir, the Broker object must now be
explicitly created. This also gives an easier way to have multiple Brokers,
if desired.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Have only one user_config_dir, so change API to enchant_get_user_config_dir,
returning a char *. Make ENCHANT_CONFIG_DIR override it, not supplement it:
this is intended for use by applications embedding Enchant which want to cut
off its external configuration, and for testing.
Look for modules only in one place, under the given prefix (which can be
overridden by ENCHANT_PREFIX_DIR).
The system enchant.ordering is moved to sysconfdir from pkglibdir. Little
point having two separate files, but make it configurable by the sysadmin.
Fix default path relative to g_get_system_data_dirs for finding hunspell
dictionaries (was still using one more suited to myspell).
Make a couple of C++ fixes to the uspell provider. Not currently being built.
Don’t look for Enchant config directly in home directory any more.
Rename some internal functions without leadiing underscore. No point doing
this for merely static functions, as we don’t with most names. Only leave
underscores where there is already a function with the name without the
underscore.
Reverse the order of directories in enchant_get_conf_dirs, so it doesn’t
have to be reversed by its only caller.
Update relevant tests.
|
|
|
|
|
|
| |
Not needed any longer for testing (use ENCHANT_PREFIX_DIR); for other
purposes it’s mostly useful for chaos (e.g. code injection into applications
using Enchant).
|
|
|
|
|
|
|
|
|
|
|
| |
I had misunderstood the libtool documentation: LT_OBJDIR is always used for
dynamic objects. (Confirmed by reading the code.)
Also, the libtool variable objdir holds the same value, so AC_SUBST it
directly rather than inventing our own variable.
This also fixes an incompatibility with Windows for the integration test
environment (path separator).
|
|
|
|
|
| |
Move READMEs from individual providers into comments at the top of the
relevant source file.
|
|
|
|
|
| |
Emacs is already dealt with via .dir-locals.el; other editors should
similarly be set up globally.
|
|
|
|
|
|
|
|
| |
It doesn’t work with current ispell (3.4) dictionaries on at least some
systems. The ispell maintainer is not interested in making ispell a
library (despite an offer of help). This seems quite reasonable.
The ispell back-end will be retired to its own repository (like Uspell).
|
| |
|
|
|
|
| |
This includes the user-visible name for the back-end.
|
|
|
|
| |
Fix a warning found by compiling in src with more warnings.
|
|
|
|
|
| |
Applies to both free and g_free, both of which explicitly specify that
passing NULL is acceptable.
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
|
|
|
| |
Much of this work was done by Hubert Figuière.
|
| |
|
|
|
|
| |
Forgot to update use of noinst_PROGRAMS to check_PROGRAMS
|
| |
|
| |
|
|
|
|
|
| |
Improve error messages, and ensure test-enchant fails if a dictionary or PWL
can’t be created.
|
|
|
|
|
|
|
|
|
|
|
| |
This applies to informal English (e.g. ‘asn’t, h-less hasn’t), and of course
some other languages, e.g. Dutch.
Add a note about deprecation of G_UNICODE_COMBINING_MARK.
Reorganise is_word_char to put code in more logical order: test for
apostrophe at the top (include curly apostrophe), as its GUnicodeType is not
needed.
|
| |
|
| |
|