[Libreoffice-commits] core.git: sc/source
LuboÅ¡ LuÅák (via logerrit)
logerrit at kemper.freedesktop.org
Thu Sep 10 09:16:30 UTC 2020
sc/source/core/data/column4.cxx | 28 ++++++++++++++++++++++++++--
1 file changed, 26 insertions(+), 2 deletions(-)
New commits:
commit 4f36f2ccab6286ec09480caea602c0fa19195736
Author: Luboš Luňák <l.lunak at collabora.com>
AuthorDate: Fri Sep 4 15:51:07 2020 +0200
Commit: Luboš Luňák <l.lunak at collabora.com>
CommitDate: Thu Sep 10 11:15:47 2020 +0200
detect if a cell still needs interpreting after Interpret()
https://bugs.documentfoundation.org/attachment.cgi?id=51878
after hard-recalc asserts because of not having cell dependencies
non-dirty for threaded calculation. This is because Interpret()
actually sometimes returns without computing a value for the cell,
e.g. when it backs out because of a need to do iteration.
This is handled when Interpret() is called from InterpretTail(),
but ScDependantsCalculator does not handle this and considers
all cells interpreted, unless it detected a problem.
We've already fixed a number of such bugs, and given that there
are still problematic corner cases after all this time, add code
simply detecting this generic problem and avoiding threading
in that case, with a SAL_WARN. This does not fix the problem
itself, but at least now it's handled.
Change-Id: I2f454b577f6516d2ce008005dbfbeb554e18d811
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/102156
Tested-by: Jenkins
Reviewed-by: Luboš Luňák <l.lunak at collabora.com>
diff --git a/sc/source/core/data/column4.cxx b/sc/source/core/data/column4.cxx
index ae3cd16fffa7..f6dac8ae3c65 100644
--- a/sc/source/core/data/column4.cxx
+++ b/sc/source/core/data/column4.cxx
@@ -1667,6 +1667,10 @@ static bool lcl_InterpretSpan(sc::formula_block::const_iterator& rSpanIter, SCRO
// Found a completely dirty sub span [nSpanStart, nSpanEnd] inside the required span [nStartOffset, nEndOffset]
bool bGroupInterpreted = pCellStart->Interpret(nSpanStart, nSpanEnd);
+ if (bGroupInterpreted)
+ for (SCROW nIdx = nSpanStart; nIdx <= nSpanEnd; ++nIdx, ++itSpanStart)
+ assert(!(*itSpanStart)->NeedsInterpret());
+
ScRecursionHelper& rRecursionHelper = rDoc.GetRecursionHelper();
// child cell's Interpret could result in calling dependency calc
// and that could detect a cycle involving mxGroup
@@ -1686,7 +1690,15 @@ static bool lcl_InterpretSpan(sc::formula_block::const_iterator& rSpanIter, SCRO
++itSpanStart;
for (SCROW nIdx = nSpanStart+1; nIdx <= nSpanEnd; ++nIdx, ++itSpanStart)
{
- (*itSpanStart)->Interpret(); // We know for sure that this cell is dirty so directly call Interpret().
+ if( !(*itSpanStart)->Interpret()) // We know for sure that this cell is dirty so directly call Interpret().
+ {
+ SAL_WARN("sc.core.formulagroup", "Internal error, cell " << (*itSpanStart)->aPos
+ << " failed running Interpret(), not allowing threading");
+ bAllowThreading = false;
+ return bAnyDirty;
+ }
+ assert(!(*itSpanStart)->NeedsInterpret());
+
// Allow early exit like above.
if ((mxParentGroup && mxParentGroup->mbPartOfCycle) || !rRecursionHelper.AreGroupsIndependent())
{
@@ -1788,7 +1800,19 @@ static void lcl_EvalDirty(sc::CellStoreType& rCells, SCROW nRow1, SCROW nRow2, S
else
{
// No formula-group here.
- bool bDirtyFlag = (*itCell)->MaybeInterpret();
+ bool bDirtyFlag = false;
+ if( (*itCell)->NeedsInterpret())
+ {
+ bDirtyFlag = true;
+ if(!(*itCell)->Interpret())
+ {
+ SAL_WARN("sc.core.formulagroup", "Internal error, cell " << (*itCell)->aPos
+ << " failed running Interpret(), not allowing threading");
+ bAllowThreading = false;
+ return;
+ }
+ assert(!(*itCell)->NeedsInterpret());
+ }
bIsDirty = bIsDirty || bDirtyFlag;
// child cell's Interpret could result in calling dependency calc
More information about the Libreoffice-commits
mailing list