Control: retitle -1 pandas: high RAM usage in test suite Control: severity -1 normal
The ones I've tried (~half the affected ones) pass on qemu-mipsel when run in smaller batches.
Memory usage for the whole test suite (on amd64) starts at nearly 1GB just for *collecting* the tests and continues to rise during the test suite run. (I don't know if this is a memory _leak_ or the normal result of such a large (~61k items) test suite.)
Given the above, I suspect the MemoryErrors are this cumulative memory use, not any single test using too much, and hence that the issue can be fixed by splitting up the test suite.
(The new failures on arm*/ppc64 are a time zone conversion test using now() as input, which these architectures happened to build during a DST transition => AmbiguousTimeError.)
