[Libreoffice-commits] core.git: Changes to 'feature/cib_contract139'

Fri Dec 21 00:48:16 UTC 2018

New branch 'feature/cib_contract139' available with the following commits:
commit b0bbfb61c68374012e1c9fb6000c166f146071e9
Author: Thorsten Behrens <Thorsten.Behrens at CIB.de>
Date:   Fri Dec 21 00:57:37 2018 +0100

    fix crash
    
    Change-Id: Ia0332b62daad878dd3af939a15a9947562b8ed0d

commit ab715cf3a1247603932d87af57e20df7d9e723ba
Author: Noel Grandin <noel.grandin at collabora.co.uk>
Date:   Thu Jul 26 14:36:38 2018 +0200

    loplugin:returnconstant in svl,svtools
    
    Change-Id: Id297a513f3313e10531f0ccd99a16277e4e37fa1
    Reviewed-on: https://gerrit.libreoffice.org/58111
    Tested-by: Jenkins
    Reviewed-by: Noel Grandin <noel.grandin at collabora.co.uk>
    (cherry picked from commit 928dfebe109794eb079469a46f549e25b3b1e89d)

commit 742c800c0f4ba82e293f866b81b57ceb44b86fa9
Author: Fridrich Štrba <fridrich.strba at bluewin.ch>
Date:   Thu Jul 12 11:57:07 2018 +0200

    call System.runFinalizersOnExit by reflection, since it was removed in jdk11
    
    Change-Id: I542c87bc1de21727a035cc6ac3b3e20c0ccfb5f7
    (cherry picked from commit 983035cc974faf9a2dcc1ecdf01391e618970a7f)

commit c249e2284e122f5b515a3a95dbe2a4f25381511f
Author: Armin Le Grand <Armin.Le.Grand at cib.de>
Date:   Thu Dec 20 17:31:32 2018 +0100

    Enhance TaggedPDF export (accessibility)
    
    The current tagged PDF export does not well support
    quite some internal structures. This includes all
    apps (Draw/Impress/Writer/Calc) and some areas.
    
    Area AlternativeText ('/Alt'):
    
    Only writer currently at least adds Title information,
    but we also have Description (MS does add) and Name.
    Target is to add this information when available to
    content frames.
    Writer did that by manually adding that tag using
    PDFExtOutDevData::SetAlternateText, but only used
    Title so far.
    To make this work as broad as possible, better add
    this to primitives. There is already a primitive called
    ObjectInfoPrimitive2D that encapsulates any content
    adding Name/Title/Description using GroupPrimitive
    functionality.
    Changed Writer to use that way. Draw/Impress already
    uses it, all apps now use graphic paint using primitives,
    so we have a natural target to encapsulate. Add support
    to VclMetafileProcessor2D to interpret it and add
    - if mpPDFExtOutDevData->GetIsExportTaggedPDF() - that
    data using a combination of Name/Title/Description and
    add using mpPDFExtOutDevData->SetAlternateText.
    This works for Draw/Impress/Writer, but not for Calc
    because Calc does not create more complex data structures,
    so SetAlternateText does not work (see
    PDFWriterImpl::setAlternateText for more infos).
    
    Area Tagged ListContent (use 'L', 'LI', 'LBody' PDF tags):
    
    To support this in Draw/Impress, we can also use a similar
    way to support in primitives. For this I evaluated how to
    add needed OutlineLevel information to the existing (and
    already used to write 'P') TextHierarchyParagraphPrimitive2D.
    Added this and now ready to use in VclMetafileProcessor2D
    ::processTextHierarchyParagraphPrimitive2D.
    Added now using the OutlineLevel information at the
    TextHierarchyParagraphPrimitive2D. Made sure there are
    fallbacks to unchanged old behaviour when no PDF export
    or no Tagged-PDF used. Creating now '/L', '/LI' and '/LBody'
    statements as tagged PDF wants us to do.
    Exported PDF still works well while additionally a verifier
    as 'PAC 3' shows the expected and wanted structure.
    This will work now for any text in Draw/Impress and for
    Draw-Objects using Lists in Calc. Need to check for direct
    text in Calc cells and Writer - and guess how big the
    effort would be for these to make it work there, too.
    
    Area '/Artifact':
    
    Target is to avoid too much ScreenReader hassle when
    Impress uses Pictures/FillPatterns etc. in Background
    - what means on MasterPage in Impress.
    Experimented with different possibilities. Decided to use
    existing StructureTagPrimitive2D and extend for info if
    encapsulated data is 'Background' data -> on MasterPage.
    Can be created in ImplRenderPaintProc in method
    createRedirectedPrimitive2DSeque as needed by checking
    for MasterPage member (remember: primitives need to be
    as independent from model data as possible, never include
    e.g. a SdrObject reference in any way).
    Tried different ways to use this in VclMetafileProcessor2D
    processStructureTagPrimitive2D, see comments there. Current
    best solution is to just *not* create StuctureTag information
    for these objects.
    
    Change-Id: Ib2a578b02c1256758cda6d15ce37799803d8205c

commit 95b00807849c22b91cdce19fde5404ccb89469a7
Author: Caolán McNamara <caolanm at redhat.com>
Date:   Tue Jul 31 15:33:43 2018 +0100

    crashtesting: stack exhaustion exporting moz1253590-2.svg to odg
    
    happens on the crashtesting box, but not locally, comparing the stack frames its clear
    that the stack used on the crashtester for VclPixelProcessor2D::processBasePrimitive2D is
    over double that used locally...
    
    comparison on
    > objdump -S workdir/CxxObject/drawinglayer/source/processor2d/vclmetafileprocessor2d.o |less
    gives...
    void VclMetafileProcessor2D::processBasePrimitive2D(const primitive2d::BasePrimitive2D& rCandidate)
    gcc-4.8.2-3.2.mga4 has...       sub    $0x5b0,%rsp
    vs...
    gcc-8.1.1-5.fc28.x86_64 has...  sub    $0x2e0,%rsp
    
    lets split up this method
    
    Change-Id: I6d84f555a01b5c58f530adb9b9b8cb8803c985bf
    Reviewed-on: https://gerrit.libreoffice.org/58364
    Tested-by: Jenkins
    Reviewed-by: Caolán McNamara <caolanm at redhat.com>
    Tested-by: Caolán McNamara <caolanm at redhat.com>
    (cherry picked from commit aa2e694e8d9e22de94dbf21f81883f9af0e34ce9)

commit b98122fe511b48f4e99bb25d6d7becd8d52d1d19
Author: Caolán McNamara <caolanm at redhat.com>
Date:   Tue Jul 31 14:20:52 2018 +0100

    crashtesting: stack exhaustion exporting moz943243-4.svg to odg
    
    happens on the crashtesting box, but not locally, comparing the stack frames its clear
    that the stack used on the crashtester for VclMetafileProcessor2D::processBasePrimitive2D is
    over double that used locally...
    
    comparison on
    > objdump -S workdir/CxxObject/drawinglayer/source/processor2d/vclmetafileprocessor2d.o |less
    gives...
    void VclMetafileProcessor2D::processBasePrimitive2D(const primitive2d::BasePrimitive2D& rCandidate)
    gcc-4.8.2-3.2.mga4 has...	sub    $0x11e8,%rsp
    vs...
    gcc-8.1.1-5.fc28.x86_64 has...	sub    $0x4c0,%rsp
    
    lets split up this method
    
    Change-Id: I73ef1eb0280224988176986918a2d025344197d0
    Reviewed-on: https://gerrit.libreoffice.org/58362
    Reviewed-by: Caolán McNamara <caolanm at redhat.com>
    Tested-by: Caolán McNamara <caolanm at redhat.com>
    (cherry picked from commit 2a4b054b3685aaeca5bdcb2969bcfc25365421d9)

commit efd4080ff840c9aedc568f38255d9cbf279f98f3
Author: Noel Grandin <noel.grandin at collabora.co.uk>
Date:   Fri Jul 27 14:15:41 2018 +0200

    loplugin:returnconstant in ucbhelper,drawinglayer
    
    Change-Id: I4e84c570fe0e555a3aeb1f11632715de466d6e0e
    Reviewed-on: https://gerrit.libreoffice.org/58192
    Tested-by: Jenkins
    Reviewed-by: Noel Grandin <noel.grandin at collabora.co.uk>
    (cherry picked from commit 53491f8a66d72176aab332983f9aff47ecd3d654)