we can just take a "const &".
(found by running clang-tidy with the
performance-unnecessary-copy-initialization warning)
Change-Id: I20fd208c65303da78170b1ac06c638fdf3aa094b
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/176267
Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
Tested-by: Jenkins
When there is a grammar check error that it has no suggestions for, Duden
may return a JSON without suggestions, but still informing about the type
of the error, like
...
{
"offset": 468,
"length": 8,
"errorcode": 21,
"type": "orth",
"proposals": []
},
...
Before the change, we underlined the error, but right-clicking it would
only bring the normal Writer's context menu, as if there were no error.
This was made in commit 7697ef9d8f (Proof
reading suggestions:, 2023-11-28); before that change, the context menu
was shown, but since it had no suggestions, and no explanation text, it
was considered better to just not show the grammar check results. But
no menu for an underlined word is no better; so let's instead show the
explanation, based on the type of the error reported by Duden - that is
stored in SingleProofreadingError::aShortComment. Also set aFullComment,
which is used in the grammar check dialog.
Incidentally, it even improves popups with suggestions, because it now
has the explanation as well, where previously only was an icon without
any text.
It may be further improved in a follow-up, because there is "errorcode"
in the message, which we may use to add more details to the explanation.
tdf#160127 should be fixed with this, too; but if there are cases when
there are neither suggestions, nor explanation available, it will need
further work.
Change-Id: I62cde204e0142d6226c9c9486f1cdb5ffc20dca8
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/175017
Reviewed-by: Mike Kaganski <mike.kaganski@collabora.com>
Tested-by: Jenkins
Add new hyphenation option "Compound characters at line end",
equivalent of libhyphen's COMPOUNDLEFTHYPHENMIN, to limit bad
pattern based hyphenation of compound words using morphological
analysis of Hunspell.
* Add checkbox to Text Flow in paragraph formatting dialog window
* Store property in paragraph model:
css::style::ParagraphProperties::ParaHyphenationCompoundMinLeadingChars
* Add ODF import/export (loext:hyphenation-compound-remain-char-count)
* Add ODF unit tests
Note: slower Hunspell based hyphenation is used only if
ParaHyphenationCompoundMinLeadingChars >= 3 (we assume that
libhyphen hyphenation patterns cover the smaller distances
correctly). Hunpell based hyphenation doesn't introduce
new hyphenation breaks, only detects the stem boundaries
from the libhyphen based hyphenation breaks.
Follow-up to commit c899d3608d
"tdf#158885 sw: don't hyphenate right after a stem boundary",
replacing hyphenation zone dependence with the new "Compound
characters at line end".
Note: preset COMPOUNDLEFTHYPHENMIN values aren't loaded yet
from hyphenation dictionaries.
Note: the suffix of the last stem of the compound is always
hyphenated, i.e. the distance limits only hyphenation
inside the stem, not inside its suffix or at the end of the
stem before the suffix.
Change-Id: I46a0288929a66f7453e3ff97fbc5a0c6a01f038f
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/164983
Tested-by: László Németh <nemeth@numbertext.org>
Reviewed-by: László Németh <nemeth@numbertext.org>
deprecated in 2008 (and seemingly never even implemented)
Replace the use of this property in sample code and
in Java unit tests with another property (I arbitrarily
selected "IsIgnoreControlCharacters")
Change-Id: I2cb474f895ac221952b59e98bfdbc01012095519
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/164355
Tested-by: Jenkins
Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
in compound words to get better typography or orthography
with more readable text, if hyphenation zone is enabled.
If there are multiple possible break points in the word
according to the libhyphen based hyphenation, keep only
the best ones using Hunspell morphological data
based on compound word decomposition of non-dictionary
words (pa: fields), and extra morphological data of dictionary
words (hy: fields) or their combination.
For readability and tradition, orthography and typography
prefer or only allow hyphenation between stems in compound
words in several languages, like Danish, Dutch, German,
Hungarian, Norwegian and Sweden.
Hyphenation zone is to avoid of too much or bad hyphenation.
Preferring stem boundaries for hyphenation within the hyphenation
zone is a natural extension of it, i.e. skip hyphenation within
stems, if there is stem boundary within the hyphenation zone.
Now skip break points after skip boundaries, if their
distance is 3 or less characters (COMPOUNDLEFTHYPHENMIN = 4).
Skip also break points on stem boundaries, if there is a
weighted stem boundary before them within 3 characters.
Weighted stem boundaries are there between the pa: fields,
(stems resulted by the compound word decomposition),
or in hy: field, boundaries marked by double || instead of
a single |.
More information: man (5) hunspell, and option -m hunspell.
Note: break points skipped only in the last stems for languages
with fogemorphemes, yet, because of their incomplete Hunspell
output for morphological analysis.
Change-Id: I739908716d11a9c2db0c9d36fba8657ba6f53bee
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/161498
Tested-by: Jenkins
Reviewed-by: László Németh <nemeth@numbertext.org>
The problem is that curl 8.3.0 removed the NSS backend, so we now
have no other choice than to use the bundled OpenSSL on Linux.
Currently any curl https connection fails with:
CurlSession.cxx:963: curl_easy_perform failed: (60) SSL certificate problem: unable to get local issuer certificate
Apparently this requires manually telling curl which CA certificates to
trust; there is a configure flag --with-ca-bundle but that is useless as
it tries to load the file relative to whatever is the current working
directory, and also did i mention that there are at least 3 different
locations where a Linux system may store its system trusted CA
certificates because ALL ABOUT CHOICE.
So add a new header with an init function to try out various file
locations listed in this nice blog article and call it from way too many
places that independently use curl.
https://www.happyassassin.net/posts/2015/01/12/a-note-about-ssltls-trusted-certificate-stores-and-platforms/
TODO: perhaps bundle a cacert.pem as a fallback in case the system chose
to innovate by putting its certificates in yet another unexpected place
(regression from commit c2930ebff8)
Change-Id: Ibf1cc0069bc2ae011ecead9a4c2b455e94b01241
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/158915
Tested-by: Jenkins
Reviewed-by: Michael Stahl <michael.stahl@allotropia.de>
Reason for revert: See discussion in https://gerrit.libreoffice.org/c/core/+/154302
- it's causing CppunitTest_sw_layoutwriter2 to fail on Windows.
- it's causing CppunitTest_sw_layoutwriter to fail on Linux.
- Probably other recent CI failures are also related.
- A similar commit was reverted one year ago for similar reasons -> https://gerrit.libreoffice.org/c/core/+/135859
This commit contains the following commits:
Revert "lok: remove old hack for LanguageTool locales"
This reverts commit 9a5329a266.
Revert "lok: LanguageTool provides list of languages"
This reverts commit 21d0489a5e.
Revert "lok: remove duplicated locales for LanguageTool"
This reverts commit a0865169ab.
Revert "Load the locales from config file for languagetool"
This reverts commit 81b0d9a951.
Change-Id: I3dc48097615f510e33e233e868b6b10704d81e67
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/154342
Tested-by: Jenkins
Reviewed-by: Xisco Fauli <xiscofauli@libreoffice.org>
For old-style hunspell dictionaries, if “ar” dictionary is found, add
the rest of Arabic locales for that dictionary’s aLocaleNames, so that
distributions don’t have to symlink the dictionary files.
This is similar to the fix for macOS spell checking in:
Author: Khaled Hosny <khaled@aliftype.com>
Date: Tue Feb 7 21:29:12 2023 +0200
tdf#153436: Improve handling of Arabic system spell dictionary on macOS
Change-Id: I96b23566cba7b345a10e10c10f1bfe02ce1f8cf3
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/152706
Tested-by: Jenkins
Reviewed-by: خالد حسني <khaled@libreoffice.org>
So that when the same text is re-checked using different
settings (e.g., language), the cached result is not used.
Change-Id: Ic4ed63c6835f9d3935f1b9541c80822bb52313c2
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/148200
Tested-by: Jenkins
Reviewed-by: Mike Kaganski <mike.kaganski@collabora.com>
Do it like LanguageToolGrammarChecker::doProofreading(), which already
checked for this setting.
Helps when soffice starts up in an environment without network access
and also is pointless. Avoids this warning:
warn:languagetool:13511:13511:lingucomponent/source/spellcheck/languagetool/languagetoolimp.cxx:508: CURL request returned with error: 28
Which looked odd, given that LT was disabled already.
Change-Id: Ideb0c59ad79bdbd7d22acef22e31a43377d30ec1
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/148166
Reviewed-by: Miklos Vajna <vmiklos@collabora.com>
Tested-by: Jenkins
The system has only a generic “ar” spell dictionary, but LibreOffice
wants country code as well. Since it is the same dictionary for all
Arabic locales, we now add entries to all Arabic country codes we know
about.
Change-Id: Iaf5a8f495b3d1023d7fe40654afa0cce3fe9fbdc
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/146641
Tested-by: Jenkins
Reviewed-by: خالد حسني <khaled@aliftype.com>
Noel Grandin