Today work was solely focused on reviewing and making extensive changes to the AIX patch.
Yesterday everything seemed alright. Indeed, that was in large part due to my unstable connection to the remote AIX machine from which I was sourcing the definitions. But today I somehow managed to get a more stable connection. Apparently, if I switch back and forth between the SFTP mountpoint and some other directory on my local filesystem, the SSH connection triggers a reconnect. I have not the slightest idea why that is but i currently don’t have the time to look into it.
With this finding, I decided to go through all commits again, ensuring my changes were correct. Alas, they were not. The changes that I mentioned yesterday have already been effected. Only unsuffixed types remain (except in the cases where the “64”-suffixed types have a different definition.) Work has also gone into moving the few types that were declared specifically in the powerpc64 module back into the top-level aix module. This decision was made after the fact there’s only a single supported Rust target with this operating system.
Then I did a bunch of cross-referencing across routines I had marked with the large file API cfg. Some of them just needed deprecation as they relied on suffixed (deprecated) types. The fact that we only ever target 64-bit environments also made me realize that the two other cfgs I had added for configuring the signedness of time64_t and for defaulting to 64-bit definitions on unsuffixed types were not necessary. They have been removed in favor of the C type definitions (which would yield non-64-bit wide types in targets whose machine word size was not 64-bits.) There were also quite a few types I had initially decided to ignore due to my crappy connection to the AIX machine. Those are being looked into now.
I’m still not satisfied with the end result. This is WIP.
None at present.
The AIX work will likely take two more days. The write up for the PR has been modified but will likleky require further changes. The file references included in the source list still need updating. Actual work on the patch will likely be done by tomorrow. Then I will have to reorganize the whole Git shebang. That is likely going to take an entire day because of the amount of mid-rebase conflicts I’m going to have to solve as I juggle commits, ammend them, extract fine-grained patches into separate commits, etc. I expect the PR to be opened in three days’ time.