The macros that do the ptrlist walking don't handle empty blocks. So, when some elements are removed from a list some list rapacking need to be done to remove the empty blocks. Failure to do this can result in crashes or data corruption. This series aims to improve the situation by: - doing the repacking at the places where some elements are removed but no repacking where done. - to compensate the added cost, mark ptrlists as dirty when some element are removed from it and only do the repacking when the list is marked as dirty. Another thing that can be done would be to check the dirty bit at every list walking but I think this should be reserved for testing/debugging. In fact, while at first agreeing with the idea of the second patch, I think less and less that it's the right approach. What are we looking after here? * Performance? - Will this bit save us a lot of list repacking? * Robustness? - What this bit bring us? Nothing without check at each list walking. Is it a good idea to sprinkle some assert() on the macros? - I would much prefer to have something that either: - can do the list walking with empty blocks, or - never leave any empty blocks. and this without making the macros heavier :) I'll try later to propose something but for now ... Luc Van Oostenryck (2): add missing PACK_PTR_LIST() mark lists to be repacked as dirty cse.c | 1 + evaluate.c | 1 + ptrlist.c | 7 +++++++ ptrlist.h | 4 +++- 4 files changed, 12 insertions(+), 1 deletion(-) -- -- To unsubscribe from this list: send the line "unsubscribe linux-sparse" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html