| Message ID | 20241020024628.22469-1-richard.weiyang@gmail.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | maple_tree: current split may result in deficient node | expand |
* Wei Yang <richard.weiyang@gmail.com> [241019 22:46]: > Here are 4 patches related to correctly split node. So I haven't worked through your last bunch of patches and I had not fully responded to v3 before v4 came out. These things take time to go through because you are changing fiddly things that may break things or cause regressions, but you keep sending a new series while another is in flight. > > Patch 1: adjust the calculation of split Split calculations exist with +/-1 to avoid jitter, but I will have to look at this in minute detail. If you screw this up, then we may lose data. > Patch 2: add a test case to check deficient split > Patch 3: the min value for mab_calc_split() seems not correct > Patch 4: during validation, we skip right most node on each level > > Wei Yang (4): > maple_tree: current split may result in deficient node > maple_tree: add a test check deficient node > maple_tree: use the correct min to calculate split > maple_tree: only root node could be deficient > > lib/maple_tree.c | 8 ++++---- > lib/test_maple_tree.c | 28 ++++++++++++++++++++++++++++ > 2 files changed, 32 insertions(+), 4 deletions(-) > > -- > 2.34.1 >
On Sun, Oct 20, 2024 at 04:34:03PM -0400, Liam R. Howlett wrote: >* Wei Yang <richard.weiyang@gmail.com> [241019 22:46]: >> Here are 4 patches related to correctly split node. > >So I haven't worked through your last bunch of patches and I had not >fully responded to v3 before v4 came out. > Sorry, I thought you have finished v3. I will check with you in later work. >These things take time to go through because you are changing fiddly >things that may break things or cause regressions, but you keep sending >a new series while another is in flight. > Thanks for you time on reviewing these. This really took efforts. I will keep those in turn later. Is this what you preferred? >> >> Patch 1: adjust the calculation of split > >Split calculations exist with +/-1 to avoid jitter, but I will have to >look at this in minute detail. > >If you screw this up, then we may lose data. > >> Patch 2: add a test case to check deficient split >> Patch 3: the min value for mab_calc_split() seems not correct >> Patch 4: during validation, we skip right most node on each level >> >> Wei Yang (4): >> maple_tree: current split may result in deficient node >> maple_tree: add a test check deficient node >> maple_tree: use the correct min to calculate split >> maple_tree: only root node could be deficient >> >> lib/maple_tree.c | 8 ++++---- >> lib/test_maple_tree.c | 28 ++++++++++++++++++++++++++++ >> 2 files changed, 32 insertions(+), 4 deletions(-) >> >> -- >> 2.34.1 >>