Looks like idlc ignores the lone (due to the preceeding ";") raises
specification, for whatever buggy reason. XAxis.idl had been like this since
the inception, and the C++ implementation of getAxisGroup in
sc/source/ui/vba/vbaaxis.{cxx,hxx} had also failed to list BasicErrorException
since its inception (likely because the function signatures were copied directly
from the generated .hdl file, which of course also lacks BasicErrorException;
and also does not really need it given the function definition just returns a
member variable), so keep the status quo and clean up the .idl file.
Change-Id: Ib4ffddb6ec867ca302cc595915cae17099734ddf
UNOIDL identifiers starting with a lowercase letter and containing underscores
are reserved for the implementation. It would be good to be able to enforce
that in code parsing UNOIDL files, but some existing identifiers violate that.
So at least change any violating method parameter identifiers here. While that
is theoretically an incompatible change for published interfaces, it practically
does not matter at all.
Change-Id: I5eff17b5dd5e2e92984184127c4fe1712d62c9dd
There were two problems here:
1) OOXML has no way to explicitly disable the footnote separator, what
is does is that it omits the <w:separator/> element in that case. We
didn't parse that previously -- now we do, and if it's missing, the
separator is disabled.
2) The footnote stream isn't read by the importer, only when the main
stream references the footnote one, the relevant part of it is parsed.
At the moment we always parse the first (special, "separator") entry in
the footnote stream, that may be optimized later if it becomes a
bottleneck.
Change-Id: Ie588270a212fc90fc41095029a362cfd832b24f8
- there was a symbol mismatch for dynamic loading
- while this fixes fdo#68210, similar fixes are likely needed for the other
dynamically loaded functionality
Change-Id: I7e4d6baf256dbe7b145dcbc118b4cac2075e07f5
...it had been deprecated at least since late OOo times, with the
css.sheet.NamedRangeFlag constant group as replacement. (UNOIDL identifiers
starting with an underscore are illegal. It would be good to be able to enforce
that in code parsing UNOIDL files, but some existing identifiers like this one
violate that.)
Change-Id: Ib8067dee47cec46356065b7b70cc6b47b97e5bc0
mv only moves non-hidden files out of UNZIP_DIR, hence removal of
UNZIP_DIR can fail if there are hidden files remaining. This assumes
that hidden files aren't actually needed for our purposes.
This is a problem e.g. for libatomic_ops which contains a .gitignore
in it's top directory, causing the removal of UNZIP_DIR to fail.
Change-Id: Ia4a621b90bc4cc5fc15dd2a3ecc209734abc6269
Reviewed-on: https://gerrit.libreoffice.org/5808
Reviewed-by: Michael Stahl <mstahl@redhat.com>
Tested-by: Michael Stahl <mstahl@redhat.com>
...doxygen still picks them up, as it traverses the complete udkapi/ and offapi/
soruce trees. (And rename udk-modules.idl to modules.idl for consistency.)
Change-Id: Ic52c333756810c285059f03edc207a0913ead160
UNOIDL identifiers starting with a lowercase letter and containing hyphens are
reserved for the implementation. It would be good to be able to enforce that in
code parsing UNOIDL files, but some existing identifiers violate that.
In the case of the published, deprecated css.uno.Uik, the change is incompatible
in theory but arguably irrelevant in practice.
Change-Id: I61f66e2d73c6aca5498ae566758893b546eb81d5
Remove an odd {1} repeat count, and remove the "."{DIGIT}+ pattern that is
already covered by the following {DIGIT}*"."{DIGIT}+ pattern.
Change-Id: If99dfe10a5e37225355472bf4deceb34c0a92eb9
To be used around code where some compiler, in some circumstances,
generates bogus warnings about unreachable code, that it would be much
uglier to work around otherwise.
Specifically, I will at first now use this to get rid of MSVC warnings
about unreachable code when calling a function defined in another
source file (but going into the same library) that always throws. The
compiler notices this when one uses link-time code generation and it
thus can do global inlining of code from all compilation units that go
into a library (or executable).
For MSVC, the __pragma that the SAL_WNOUNREACHABLE_CODE_PUSH macro
expands to needs to be in force at the start curly brace of a
function, so place the PUSH macro before the function definition. For
clarity, I guess it is best that the corresponding POP macro comes
after the end of the function.
Change-Id: Icef5259c5360b9facdc136fec1f207665ce79d90
That should be older than what RHEL-5 have, so closer to our real
baseline. Cleanup checks for older FreeType as a side effect.
Change-Id: I10dc6a3064b6be88cd7bdc72c501d50df4a2613b
Reviewed-on: https://gerrit.libreoffice.org/5804
Reviewed-by: Caolán McNamara <caolanm@redhat.com>
Tested-by: Caolán McNamara <caolanm@redhat.com>
We require at least 2.0.9 right now anyway, so no drop the dlsym hacks.
This is essentially a revert of a36dab4a59
from 2003.
Change-Id: Id9bca2962f25d92f1707bc3d1efc054ffd22c714
Reviewed-on: https://gerrit.libreoffice.org/5803
Reviewed-by: Caolán McNamara <caolanm@redhat.com>
Tested-by: Caolán McNamara <caolanm@redhat.com>
Build Catalan-Valencian as ca-valencia instead of ca-XV private-use.
Introduced LANGUAGE_CATALAN_VALENCIAN 0x0803 mapping to ca-ES-valencia,
preserving old ca-XV and qcv-ES mappings to now
LANGUAGE_CATALAN_VALENCIAN and LANGUAGE_OBSOLETE_USER_CATALAN_VALENCIAN
0x8003 to ca-ES-valencia.
Removed special !bUserInterfaceSelection treatment from
MsLangId::getReplacementForObsoleteLanguage() and added the usual
obsolete replacement instead.
Change-Id: I2fdd8b0bac55d4b4ae2cbf3c3645f09fefec9b6e
Black and white color option found in Impress at
File-->Print-->Libreoffice Impress tab
Before change, images are printed in pure white, and hence
missing from the printouts.
Should resolve bug 62838, and is contributed under the LGPLv3+ / MPL.
Change-Id: I29e291c2f2baf95b5725b862625f6a206e4cb8ba
Reviewed-on: https://gerrit.libreoffice.org/5798
Reviewed-by: Caolán McNamara <caolanm@redhat.com>
Tested-by: Caolán McNamara <caolanm@redhat.com>
This change has 2 purposes:
* fix gradient background when writing pdf. Before a gradient filled
frame would be displayed as white (or transparent), because the transp.
rendering mode was taking precedence over the gradient one.
* making the code more readable by grouping multiple conditions in
an enum.
Change-Id: I5a30756b72be5eabf364c67e175c7b9d22bfa1c0
Reviewed-on: https://gerrit.libreoffice.org/5759
Reviewed-by: Miklos Vajna <vmiklos@collabora.co.uk>
Tested-by: Miklos Vajna <vmiklos@collabora.co.uk>
Noticed during inlining at link-time code generation phase: The called
function always throws. So let's hardcode the throw for now then at
the callsite instead. Not ideal, I know. Add comments describing what
is going on.
Change-Id: I60d14b25aa62846fc0314aad7e00b2990f4cff26
