[Libreoffice-commits] core.git: drawinglayer/source

Caolán McNamara caolanm at redhat.com
Fri Oct 24 07:18:30 PDT 2014


 drawinglayer/source/processor2d/vclhelperbufferdevice.cxx |    5 +++++
 1 file changed, 5 insertions(+)

New commits:
commit e25a020d59b019893d2e04ac61e4ed25ef0a6e61
Author: Caolán McNamara <caolanm at redhat.com>
Date:   Fri Oct 24 15:10:15 2014 +0100

    Resolves: fdo#84729 buffer cache grows to over 9000 unused entries
    
    regression from
    
    commit a5351434fbea1126961f5dbaa986e7c9e686c24e
    Author: tsahi glik <tsahi.glik at cloudon.com>
    Date:   Thu Jan 16 12:51:44 2014 -0800
        fix rendering issues in iOS with aqua
    
    because now the cache just fills up because nothing will have
    a depth of "0" and a new entry is created every time.
    
    0 could be seen to mean "whatever is a good non-1bit depth" or auto or
    something. (Do the uses of "8" really mean 8)
    
    0 passed to new VirtualDevice under Linux/X ends up using the depth of the
    outputdev, so here set nBits to that value to start with and the problem
    goes away
    
    Change-Id: I4390d6f7edef9dc7430cc77e3518dc751a5f479d

diff --git a/drawinglayer/source/processor2d/vclhelperbufferdevice.cxx b/drawinglayer/source/processor2d/vclhelperbufferdevice.cxx
index a3bdbf6..94a155d 100644
--- a/drawinglayer/source/processor2d/vclhelperbufferdevice.cxx
+++ b/drawinglayer/source/processor2d/vclhelperbufferdevice.cxx
@@ -85,6 +85,9 @@ namespace
         ::osl::MutexGuard aGuard(m_aMutex);
         VirtualDevice* pRetval = 0;
 
+        if (nBits == 0)
+            nBits = rOutDev.GetBitCount();
+
         if(!maFreeBuffers.empty())
         {
             bool bOkay(false);
@@ -183,6 +186,8 @@ namespace
 
         maUsedBuffers.erase(aUsedFound);
         maFreeBuffers.push_back(&rDevice);
+        SAL_WARN_IF(maFreeBuffers.size() > 1000, "drawinglayer", "excessive cached buffers, "
+            << maFreeBuffers.size() << " entries!");
         Start();
     }
 


More information about the Libreoffice-commits mailing list