Em dom., 14 de abr. de 2024 às 21:29, Michael Paquier <mich...@paquier.xyz> escreveu:
> On Sat, Apr 13, 2024 at 10:40:35AM -0300, Ranier Vilela wrote: > > This sounds a little confusing to me. > > Is the project policy to *tolerate* dereferencing NULL pointers? > > If this is the case, no problem, using Assert would serve the purpose of > > protecting against these errors well. > > In most cases, it does not matter to have an assertion for a code path > that's going to crash a few lines later. The result is the same: the > code will crash and inform about the failure. > > > rbtxn_get_toptxn already calls rbtxn_is_subtx, > > which adds a little unnecessary overhead. > > I made a v1 of your patch, modifying this, please ignore it if you don't > > agree. > > c55040ccd017 has been added in v14~, so this is material for 18~ at > this stage. If you want to move on with these changes, I'd suggest to > add a CF entry. > I think it's worth it, because it's a case of a possible bug, but very unlikely, and Heikki's suggestions improve the code. best regards, Ranier Vilela