Age | Commit message (Collapse) | Author |
|
Change-Id: I300d14d580d450ec338129918955651b9d40d5d2
Reviewed-on: https://gerrit.libreoffice.org/78059
Tested-by: Jenkins
Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
|
|
Change-Id: I5db25074d42fa22f3e36969cb561ad64c1e96dc1
Reviewed-on: https://gerrit.libreoffice.org/77760
Tested-by: Jenkins
Reviewed-by: Julien Nabet <serval2412@yahoo.fr>
|
|
Triggering those SAL_WARN sporadically caused deadlocks at least for
<https://ci.libreoffice.org/job/lo_ubsan/>.
Change-Id: I7b7037e411c29eea26e63f71a5679127b084f447
Reviewed-on: https://gerrit.libreoffice.org/77374
Tested-by: Jenkins
Reviewed-by: Stephan Bergmann <sbergman@redhat.com>
|
|
This is a partial revert of 17642437fe0d68cf868ab430f04b4fdc12f1767f "reduce
ifdef forest". The original code used unsetenv only for certain platforms, and
putenv for others, but code a few lines further down uses unsetenv
unconditionally, so assume that it is safe to use on all relevant platforms
these days.
unsetenv isn't listed as async-signal-safe at <http://pubs.opengroup.org/
onlinepubs/9699919799/functions/V2_chap02.html#tag_15_04_03> "Signal Actions"
(but is already used a few lines further down, also between fork and exec), but
at least we get rid of the memory management involved in the OUString instance.
Change-Id: Iac993db8819d40a0841c455ed04ff9ca2ee2e4eb
Reviewed-on: https://gerrit.libreoffice.org/77368
Tested-by: Jenkins
Reviewed-by: Stephan Bergmann <sbergman@redhat.com>
|
|
Change-Id: I8b9d8f80bbf3d7d9c060d2da4675ae832966c085
Reviewed-on: https://gerrit.libreoffice.org/77242
Tested-by: Jenkins
Reviewed-by: Caolán McNamara <caolanm@redhat.com>
Tested-by: Caolán McNamara <caolanm@redhat.com>
|
|
Change-Id: Icb9d9e1cd21e2506e36fe40a3b93b6a2521a868c
Reviewed-on: https://gerrit.libreoffice.org/77239
Tested-by: Jenkins
Reviewed-by: Stephan Bergmann <sbergman@redhat.com>
|
|
so drop the first write in favor of the second
Change-Id: Iac906d806a66e010e8352139b555aef6078bda02
Reviewed-on: https://gerrit.libreoffice.org/77235
Tested-by: Jenkins
Reviewed-by: Caolán McNamara <caolanm@redhat.com>
Tested-by: Caolán McNamara <caolanm@redhat.com>
|
|
Change-Id: Ie183c445bf8a545f59aac7b0e29f72ab679a6cf3
Reviewed-on: https://gerrit.libreoffice.org/76852
Tested-by: Jenkins
Reviewed-by: Andrea Gelmini <andrea.gelmini@gelma.net>
|
|
after commit bf6bde2da134dad60ecbf8f3e97674abadb7349e
Change-Id: Iaf6f1457f95e5b821879b735c16a692cca7515db
Reviewed-on: https://gerrit.libreoffice.org/76791
Reviewed-by: Mike Kaganski <mike.kaganski@collabora.com>
Tested-by: Mike Kaganski <mike.kaganski@collabora.com>
|
|
Static initializers were used for the environment strings since
commit d19c40f45dc8e8bcd9db4c6b83bdcf6367f6fbe7 to workaround
thread-unsafe getenv. The special case for Android was initially
introduced in commit 60628799633ffde502cb105b98d3f254f93115aa,
to allow modifying the environment in the code; then was fixed
in commit 4fbf6df784529d48cf194a2d9c495ffb47933d59. That was
relying on the functions being called each time their results
were used. But commit 9f027559557cb132835d8a13cdc0281ad4e757ae
changed that, making the results static, thus only calling the
functions once on all platforms.
This was effective more than a year already, so presumably the
special-casing for Android isn't needed anymore. Thus, this patch
unifies all platforms to use the same approach (calling getenv
once, storing copies of result in static variables.
Also this stores flags used in maybeOutputTimestamp to static
variables, too, to avoid parsing strings each time the function
is called.
Change-Id: I84bdbfb900c15b407fb555296d2613bd3d62da7e
Reviewed-on: https://gerrit.libreoffice.org/76573
Tested-by: Jenkins
Reviewed-by: Mike Kaganski <mike.kaganski@collabora.com>
|
|
Change-Id: I61e7a28c5beab8f0c619767545009cefae6ddcd9
Reviewed-on: https://gerrit.libreoffice.org/76772
Tested-by: Jenkins
Reviewed-by: Mike Kaganski <mike.kaganski@collabora.com>
|
|
Change-Id: Id9d994343d10b5d5e852b10946c036dfbeb66d04
Reviewed-on: https://gerrit.libreoffice.org/76656
Tested-by: Jenkins
Reviewed-by: Stephan Bergmann <sbergman@redhat.com>
|
|
It passed "make check"
Change-Id: I50c3639ec258c7fb6d73104f8e777f96995aa8e3
Reviewed-on: https://gerrit.libreoffice.org/76543
Tested-by: Jenkins
Reviewed-by: Julien Nabet <serval2412@yahoo.fr>
|
|
"an URI", to complete:
https://gerrit.libreoffice.org/#/c/75985/
Change-Id: I57489b05117fd12ae6aa22544437ab5bc6b5154f
Reviewed-on: https://gerrit.libreoffice.org/76037
Tested-by: Jenkins
Reviewed-by: Andrea Gelmini <andrea.gelmini@gelma.net>
|
|
Change-Id: I0ff36c58bf2448bdccc239582ba24b69c7431c6d
Reviewed-on: https://gerrit.libreoffice.org/75921
Tested-by: Jenkins
Reviewed-by: Caolán McNamara <caolanm@redhat.com>
Tested-by: Caolán McNamara <caolanm@redhat.com>
|
|
Let us see what happens if we annotate our mutex code,
https://scan.coverity.com/models
Change-Id: I7baf44d1a252f19b4ae47f3a6b318f7ccd9629d7
Reviewed-on: https://gerrit.libreoffice.org/75851
Tested-by: Jenkins
Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
|
|
Change-Id: Id1b8044126e65e67b2496cf7a4eb86b54ba6c1df
Reviewed-on: https://gerrit.libreoffice.org/75872
Tested-by: Jenkins
Reviewed-by: Caolán McNamara <caolanm@redhat.com>
Tested-by: Caolán McNamara <caolanm@redhat.com>
|
|
Make OffTime static const; don't cast from FILETIME to __int64 (see
https://docs.microsoft.com/en-us/windows/win32/api/minwinbase/ns-minwinbase-filetime
for explanation: "it can cause alignment faults on 64-bit Windows").
Instead, cast in opposite direction: from 8-byte-aligned 64-bit integer
to FILETIME.
Change-Id: Iba61cc0198f8f25ef471d87e661c8801724b913d
Reviewed-on: https://gerrit.libreoffice.org/75256
Tested-by: Jenkins
Reviewed-by: Mike Kaganski <mike.kaganski@collabora.com>
|
|
WPXSvInputStreamImpl was hammering on getFilePos pretty hard, and
getFilePos uses a mutex, which is slow when it is called from every
single read. So switch to using std::atomic to access position.
This is specifically fixing the performance of queryTypeByDescriptor
when called from a basic macro on a local test file.
This takes my test macro from 8s to 4s.
Change-Id: Iab707a374359e2ee0e92425b2d9a903d67cb53d4
Reviewed-on: https://gerrit.libreoffice.org/73377
Tested-by: Jenkins
Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
|
|
getpwuid_r() returns nullptr. Oddly enough, it does work on actual
iOS, though. So use hand-crafted values that match behaviour on actual
iOS.
Change-Id: Idcc95d330a93495938520229e039f340876c3653
|
|
not before vowels with a consonant sound so its a url not an url
Change-Id: Ic27ff3bee67469284d460c31ced6f63cb3633db2
Reviewed-on: https://gerrit.libreoffice.org/72062
Reviewed-by: Jens Carl <j.carl43@gmx.de>
Reviewed-by: Caolán McNamara <caolanm@redhat.com>
Tested-by: Caolán McNamara <caolanm@redhat.com>
|
|
V595 The 'ustrFile' pointer was utilized before it was verified against nullptr.
Check lines: 83, 107.
Change-Id: I874983508fb805b1f0a55b3a5b6d4d78ae1babdb
Reviewed-on: https://gerrit.libreoffice.org/71472
Tested-by: Jenkins
Reviewed-by: Mike Kaganski <mike.kaganski@collabora.com>
|
|
macosx_getLocale (via getProcessLocale, both sal/osl/unx/osxlocale.cxx) obtains
from the system a locale string like "en-DE". (Whether that is even a sane and
recommended way to obtain the system locale on macOS I don't know; but lets
leave it at that for now.)
However, setting a locale env var (LANG, LC_ALL) to such a value causes a
constructor call std::locale("") to throw a std::runtime_error
"collate_byname<char>::collate_byname failed to construct for ", at least on
macOS 10.14.4. And libdivvun (which might be bundled with a LO extension) is
known to be hit by that, see <https://github.com/divvun/libdivvun/issues/28>
"locale("") gives 'collate_byname<char>::collate_byname failed to construct
for ' on LO on mac".
The code setting LC_ALL/LC_CTYPE/LANG was there ever since
8737d1831b48acd8a4793c4728ad8563f77b5bf8 "INTEGRATION:
CWS geordi2q14: #111934#: merge CWS ooo111fix2", but for unclear reasons. Lets
assume that it had no purpose (any longer).
Change-Id: I0b519ad567a713d61f662aa984791db1a91c708c
Reviewed-on: https://gerrit.libreoffice.org/70918
Tested-by: Jenkins
Reviewed-by: Stephan Bergmann <sbergman@redhat.com>
|
|
... use GetModuleFileNameW() for that.
We call SetDllDirectoryW and SetSearchPathMode to improve security of
the process, and exclude some paths (including current directory) from
search when using API that looks for binaries whose names are not fully
qualified.
So the sequence is this:
1. Program is started using relative executable path like
"instdir\program\soffice";
2. sal_detail_initialize is called, which calls the two mentioned
hardening functions;
3. sal_detail_initialize calls osl_setCommandArgs, which in turn calls
osl_createCommandArgs_Impl;
4. The latter uses SearchPathW with empty path and first program arg
"instdir\program\soffice" to find fully qualified program path.
That now naturally fails, because current path is not searched.
But to find the process name, we need no search: we can simply use
GetModuleFileNameW() with NULL passed as module handle. Let's use that.
Note that we can't use _wpgmptr/_get_wpgmptr, because we don't use wide
entry point like wmain.
Change-Id: I7a0013a0505f7bdd38164b09b045bfade9686664
Reviewed-on: https://gerrit.libreoffice.org/70844
Reviewed-by: Mike Kaganski <mike.kaganski@collabora.com>
Tested-by: Mike Kaganski <mike.kaganski@collabora.com>
|
|
to find stuff like
OUString s = OUString("xxx")
Change-Id: Ie7ed074c1ae012734c67a2a89c564c1900a4ab04
Reviewed-on: https://gerrit.libreoffice.org/70697
Tested-by: Jenkins
Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
|
|
Change-Id: I7bd1511a6acc105ab5b42c698c7578cfb9ce06b4
Reviewed-on: https://gerrit.libreoffice.org/70708
Tested-by: Jenkins
Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
|
|
Change-Id: I7d85cbc9105c5e0c4a8d9a69c4ac9d6dfc07eabd
Reviewed-on: https://gerrit.libreoffice.org/70663
Tested-by: Jenkins
Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
|
|
We don't need the dynamic load complexity for these now with baseline
Windows version being Windows 7 SP1. Stuff used only for compatibility
with older versions was dumped.
Change-Id: I810f271796cfd875cfa18a3081c9ad444fe57b3e
Reviewed-on: https://gerrit.libreoffice.org/70321
Tested-by: Jenkins
Reviewed-by: Mike Kaganski <mike.kaganski@collabora.com>
|
|
to be declared as const. However, on platforms not excluded by
NO_PTHREAD_PRIORITY, osl_thread_priority_init_Impl() then tries
later to assign values to these. This leads to compiler errors
on platforms so affected, including FreeBSD. Revert as necessary
on affected platforms, also using NO_PTHREAD_PRIORITY to match
the code in osl_thread_priority_init_Impl().
Change-Id: I0e968231e7c1be9771844222a7bab1f0fcb51a0e
Reviewed-on: https://gerrit.libreoffice.org/69603
Tested-by: Jenkins
Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
|
|
V522 There might be dereferencing of a potential null pointer.
Change-Id: Ie617b41a8f8d334022cf5313b242a236baedba48
Reviewed-on: https://gerrit.libreoffice.org/70017
Tested-by: Jenkins
Reviewed-by: Mike Kaganski <mike.kaganski@collabora.com>
|
|
V575 The potential null pointer is passed into 'foo' function
Add asserts to those cases that are related to OOM cases. There's
nothing to be done if the assertions fail anyway.
Change-Id: I92ac95d44f512aa1948b1552b0e1f6da695a9f92
Reviewed-on: https://gerrit.libreoffice.org/70008
Tested-by: Jenkins
Reviewed-by: Mike Kaganski <mike.kaganski@collabora.com>
|
|
Change-Id: Ica146899a5557732babf4e07b20b9e3d32426c6e
Reviewed-on: https://gerrit.libreoffice.org/69824
Tested-by: Jenkins
Reviewed-by: Tor Lillqvist <tml@collabora.com>
|
|
Change-Id: I07604028845c49cc084927e21db7f21c5d053bab
Reviewed-on: https://gerrit.libreoffice.org/69796
Tested-by: Jenkins
Reviewed-by: Stephan Bergmann <sbergman@redhat.com>
|
|
The gethostbyname*() functions are obsolete.
Change-Id: I14a55eba3f111a3280f23955ffd86843079c7e75
Reviewed-on: https://gerrit.libreoffice.org/69337
Tested-by: Jenkins
Reviewed-by: Stephan Bergmann <sbergman@redhat.com>
|
|
#0 0x00007f8dc781e4dc in __lll_lock_wait () from /lib64/libpthread.so.0
#1 0x00007f8dc781a1b8 in _L_lock_820 () from /lib64/libpthread.so.0
#2 0x00007f8dc781a088 in pthread_mutex_lock () from /lib64/libpthread.so.0
#3 0x00007f8dc81d57d1 in lookupThreadId (hThread=140246229497600) at sal/osl/unx/thread.cxx:585
#4 0x00007f8dc81d5aa4 in osl_getThreadIdentifier (Thread=0x0) at sal/osl/unx/thread.cxx:701
#5 0x00007f8dc8173fa0 in osl::Thread::getCurrentIdentifier () at include/osl/thread.hxx:138
#6 0x00007f8dc81736ba in sal_detail_log (level=SAL_DETAIL_LOG_LEVEL_WARN, area=0x7f8dc81f2e53 "sal.osl",
where=0x7f8dc81f3130 "sal/osl/unx/process.cxx:238: ", message=0x5bd5b90 "Failed to exec: ENOENT", backtraceDepth=0)
at sal/osl/all/log.cxx:279
#7 0x00007f8dc816d166 in sal::detail::log (level=SAL_DETAIL_LOG_LEVEL_WARN, area=0x7f8dc81f2e53 "sal.osl",
where=0x7f8dc81f3130 "sal/osl/unx/process.cxx:238: ", stream=..., backtraceDepth=0)
at include/sal/log.hxx:50
#8 0x00007f8dc81c1ddc in ChildStatusProc (pData=0x7f8d9e302a90) at sal/osl/unx/process.cxx:238
#9 0x00007f8dc81d49a6 in osl_thread_start_Impl (pData=0x6757780) at sal/osl/unx/thread.cxx:235
#10 0x00007f8dc7817f6b in start_thread () from /lib64/libpthread.so.0
#11 0x00007f8dc7b227ed in clone () from /lib64/libc.so.6
Change-Id: I611cf907f89c7a1d8cdc3a9c42f449e8daaa2614
Reviewed-on: https://gerrit.libreoffice.org/69363
Tested-by: Jenkins
Reviewed-by: Caolán McNamara <caolanm@redhat.com>
Tested-by: Caolán McNamara <caolanm@redhat.com>
|
|
Explanation of the patch: in the Windows version of LibreOffice,
when calling the basic Shell(...) function (located in
basic/source/runtime/methods.cxx), a function is_batch_file()
is called in turn, to check if the program file path to execute
is a batch or an executable.
The function is_batch_file() is located in sal/osl/w32/procimpl.cxx
and simply checks if the file extension is equal to bat (or cmd or
btm).
This works as expected, except when the file path contains space
characters. In that case, the file path is *quoted* before the
call to is_batch_file() and for a batch file the file extension
becomes bat" (note the quote) instead of bat, and thus the call
to is_batch_file() wrongly returns false in that case.
In this patch, the issue is fixed by changing the function
get_file_extension() to make it return the correct extension
(i.e. without the '"' character) when the file name is quoted.
Change-Id: Ib6e74da87b23d64db925c17f8a26617f1a86a83d
Reviewed-on: https://gerrit.libreoffice.org/69230
Tested-by: Jenkins
Reviewed-by: Stephan Bergmann <sbergman@redhat.com>
|
|
detect static variables that can be made const.
Thanks to mike kaganski for suggesting this.
Here I introduce a new plugin feature - using markers
in nearby comments to disable the plugin for specific
vars.
Some of this stuff was old debugging code. I removed the stuff
that was older than 5 years.
Change-Id: I6ec7742a7fdadf28fd128b592fcdf6da8257585c
Reviewed-on: https://gerrit.libreoffice.org/68807
Tested-by: Jenkins
Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
|
|
Change-Id: I304621018cb1e2a47e478e86df4229bcf2176741
Reviewed-on: https://gerrit.libreoffice.org/68757
Tested-by: Jenkins
Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
|
|
This reverts commit c9bb48386bad7d2a40e6958883328145ae439cad,
and adds a bunch more fixes.
Change-Id: Ib584d302a73125528eba85fa1e722cb6fc41538a
Reviewed-on: https://gerrit.libreoffice.org/68680
Tested-by: Jenkins
Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
|
|
This reverts commit 9865440d217d975206a3f91612f0666312bc8fd8.
This is not ready to land yet, seems like the latest update
of the logic reveals a bunch more places I need to fix before it can land.
|
|
verify that parameters use the exact same typedef-names (if any)
in definition and declaration
Change-Id: I55d2817f599b0253904dce2d35a1a93967e15a77
Reviewed-on: https://gerrit.libreoffice.org/68439
Tested-by: Jenkins
Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
|
|
Change-Id: I7485ab11e4bb60420153ac53784057e5cc905c39
Reviewed-on: https://gerrit.libreoffice.org/68499
Tested-by: Jenkins
Reviewed-by: Luboš Luňák <l.lunak@collabora.com>
|
|
Change-Id: I94689e4eed290b4505d2caba2d9802ef7fb6cffd
Reviewed-on: https://gerrit.libreoffice.org/68378
Tested-by: Jenkins
Reviewed-by: Stephan Bergmann <sbergman@redhat.com>
|
|
Change-Id: Ia10e08c04fb6dfc18597fcfba3fe52cf0ba7184c
Reviewed-on: https://gerrit.libreoffice.org/68377
Tested-by: Jenkins
Reviewed-by: Stephan Bergmann <sbergman@redhat.com>
|
|
Change-Id: I7e9a709fe36b38ee8842dd34ab06f9608f498ec8
Reviewed-on: https://gerrit.libreoffice.org/68197
Tested-by: Jenkins
Reviewed-by: Stephan Bergmann <sbergman@redhat.com>
|
|
Change-Id: Ia4814242cc20fbfe35eaff95ddd64dd94e6d69af
Reviewed-on: https://gerrit.libreoffice.org/68196
Tested-by: Jenkins
Reviewed-by: Stephan Bergmann <sbergman@redhat.com>
|
|
* Some .m/.mm files that still contained tabs instead of spaces have been
cleaned up with Emacs' untabify (and
apple_remote/source/HIDRemoteControlDevice.m needed further manual adaptions):
apple_remote/source/GlobalKeyboardDevice.m
apple_remote/source/HIDRemoteControlDevice.m
apple_remote/source/KeyspanFrontRowControl.m
apple_remote/source/RemoteControl.m
vcl/osx/a11yrolehelper.mm
* Some of the changes predate 0626e66d761de18f62e4d00d427903032da9d517 "Avoid
loplugin:indentation after preproc conditional inclusion lines" and would
likely have no longer been flagged since.
Change-Id: Ibf5faffa743c7f79b36109d9879eb79d63c8c40f
Reviewed-on: https://gerrit.libreoffice.org/68090
Tested-by: Jenkins
Reviewed-by: Stephan Bergmann <sbergman@redhat.com>
|
|
Change-Id: I4b877751818febaec8e64018335dca691a476a43
Reviewed-on: https://gerrit.libreoffice.org/67561
Tested-by: Jenkins
Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
|
|
Change-Id: I82b8c49fcbbec161bf968573e28992fa5737b45b
Reviewed-on: https://gerrit.libreoffice.org/67508
Tested-by: Jenkins
Reviewed-by: Stephan Bergmann <sbergman@redhat.com>
|
|
Change-Id: I19b0cc48b18ddda5bd09c88f03565e9b73ed776a
Reviewed-on: https://gerrit.libreoffice.org/67495
Tested-by: Jenkins
Reviewed-by: Stephan Bergmann <sbergman@redhat.com>
|