https://bugs.documentfoundation.org/show_bug.cgi?id=172651

Michael Weghorn <[email protected]> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
     Ever confirmed|0                           |1
             Status|UNCONFIRMED                 |ASSIGNED
                 CC|                            |[email protected],
                   |                            |[email protected],
                   |                            |[email protected],
                   |                            |[email protected]
           Keywords|                            |bibisected, bisected,
                   |                            |regression

--- Comment #1 from Michael Weghorn <[email protected]> ---
Regression from:

    commit 7b5ddd931146d4e453134557c8bdb7a4cab3505c
    Author: Andras Timar
    Date:   Tue Apr 21 07:05:26 2026 +0200

        vcl: strip alpha in Bitmap::Convert(N24Bit) for 32-bit-with-alpha
sources

        Convert(BmpConversion::N24Bit) returned true without touching the
        bitmap when the source bit count was already >= 24, so a
        32-bit-with-alpha bitmap kept its alpha channel. Callers that rely on
        Convert(N24Bit) to strip alpha (e.g. the PDF/A transparency-removal
        path in vcl/source/outdev/transparent.cxx) then handed an alpha
        bitmap to MetaBmpScaleAction, tripping its
        !HasAlpha() assertion in dbgutil builds.

        Triggered in CPLinux-LOKit builds (--enable-cairo-rgba,
--disable-poppler)
        where pdfium-rendered PDF imports produce N32BitTcRgba bitmaps; on
        PDF/A-1 re-export the transparency-removal path then crashes in
        testTdf136805.

        Force re-encode via ImplConvertUp(N24_BPP) when HasAlpha() is set so
        the alpha channel is actually dropped.

        Signed-off-by: Andras Timar <[email protected]>
        Change-Id: I7320c5f5a873b23d4413e3183cd6ff45d9e26e08
        Reviewed-on: https://gerrit.collaboraoffice.com/c/online/+/1233
        Reviewed-by: Noel Grandin
        Tested-by: Jenkins CPCI <[email protected]>
        (cherry picked from commit cba855cd11cda78531d486279330a73133ac7151)
        Reviewed-on: https://gerrit.libreoffice.org/c/core/+/207026
        Reviewed-by: Julien Nabet
        Tested-by: Jenkins

Unless there are any concerns/other ideas (please let me know in that case),
I'd go ahead and revert that commit. The commit message sounds like it was
primarily relevant for Collabora Online, which isn't using the LibreOffice repo
anymore.

Pending revert:
https://gerrit.libreoffice.org/c/core/+/207560

-- 
You are receiving this mail because:
You are the assignee for the bug.

Reply via email to