On Wed, 13 Mar 2024 07:53:30 GMT, Christoph Langer <clan...@openjdk.org> wrote:
> 4f336085d1098e7fba7b58f0a73c028179a2a13d > ([JDK-8326718](https://bugs.openjdk.org/browse/JDK-8326718)) added a few > cases to test java/util/Formatter/Padding.java with huge Strings as > arguments. Since all possible argument combinations for the test are stored > in one array, nothing can be garbage collected while the test is running and > the heap requirement is blown up. > > In one of our test pipelines we run tier1 tests with VMs that default to 384M > of heap and this is not sufficient any longer. > > I'm improving this by splitting the one large @ParameterizedTest into > multiple ones. With that, I could run the test successfully in a test VM with > 96M of heap, e.g. by modifying `@run junit Padding` to `@run junit/othervm > -Xmx96m Padding` What about factoring out the 4 invocations of `tenMillionBlanks()` in each source method in a local var? ------------- PR Comment: https://git.openjdk.org/jdk/pull/18264#issuecomment-1993965369