On Mon, 4 May 2026 13:04:51 GMT, Erik Joelsson <[email protected]> wrote:
>>> The goal is to reduce this to one call when everything in g1 is fixed.
>>
>> Then, could we do this PR (in its "one call" patch version) after
>> "everything in g1 is fixed"?
>
> From what I can tell, the "filename" argument is a filter, used with
> `$(filter pattern…,text)` and that make function takes a list of patterns as
> input, so I can't see why it wouldn't work.
Can we not create an ExtendFlags with a `filter-out` filter?
And can simply remove files as we incrementally fix the errors.
So we end up with something like this:
$(call ExtendFlagsExcept, $(TOPDIR)/src/hotspot/share/gc/g1, %.cpp , \
g1Analytics.cpp \
g1BarrierSet.cpp \
g1BarrierSetRuntime.cpp \
g1BatchedTask.cpp \
g1BlockOffsetTable.cpp \
g1CardSet.cpp \
g1CardTable.cpp \
g1CardTableClaimTable.cpp \
g1CodeRootSet.cpp \
g1CollectedHeap.cpp \
g1CollectionSetCandidates.cpp \
g1ConcurrentMark.cpp \
g1ConcurrentRefine.cpp \
g1ConcurrentRefineThreadsNeeded.cpp \
g1EvacStats.cpp \
g1FullGCScope.cpp \
g1GCPhaseTimes.cpp \
g1HeapRegion.cpp \
g1HeapRegionEventSender.cpp \
g1HeapRegionManager.cpp \
g1HeapRegionTracer.cpp \
g1HeapSizingPolicy.cpp \
g1IHOPControl.cpp \
g1MonitoringSupport.cpp \
g1MonotonicArenaFreeMemoryTask.cpp \
g1NUMA.cpp \
g1NUMAStats.cpp \
g1Policy.cpp \
g1RemSet.cpp \
g1RootProcessor.cpp \
g1SurvRateGroup.cpp \
g1Trace.cpp \
g1VMOperations.cpp \
g1YoungCollector.cpp \
g1YoungGCPostEvacuateTasks.cpp \
, _CXXFLAGS, $(CFLAGS_CONVERSION_WARNINGS)), \
Also one unfortunate thing with with this ExtendFlags* / make warnings solution
is that we will be required to bump `MAX_PARAMS` every time we create a new
file which is caught in the filter, or remove a filtered out file. (Unless
someone has forgotten to reduce it in an earlier commit which added a file).
But it is probably a worth-while compromise for incrementally getting better
warnings.
-------------
PR Review Comment: https://git.openjdk.org/jdk/pull/30924#discussion_r3233876247
