Today efforts were centered around continuing yesterdy’s uClibc file offset bit-width transitions, and attempting to build a toolchain with which to test those changes in MIPS targets.
Yesterday’s efforts have gone on to complete half of the regex search results from yesterday’s logged search for the symbols __USE_FILE_OFFSET64 and __USE_LARGEFILE64. This is going well, and the changes made thus far have been committed to a public branch of my fork of the libc crate. No further comments will be made here because there were quite a few changes involving restructuring different records and deprecating some routines. All of this will be part of the write up that will accompany the PR once it’s open. Suffice to say there were a bunch of type definitions that were either redundant (appearing in both parent and child modules) or were otherwise wrong (and often lacking full record fields.) Much like yesterday, some of the changes instead attempt to fully transition the (possibly) 32-bit file offset types (such as off_t or ino_t) to their LFS64 variants by default. See yesterday’s report for a justification on this.
Beyond that, work on fixing the linking issues in the toolchain built with OpenADK has altogether stopped. In my search for alternative solutions, multiple other tools were found and currently efforts are centered around the Buildroot tool. This seems to provide similar functionality to OpenADK, but is (solely judging from upstream activity and the docs) more active/supported. other alternatives like the bootlin repo with toolchains were also found, but have so far not been used as they require manually setting up the shipped toolchain in an emulated environment. they also don’t seem to ship with actual images one can emulate, which is the whole point here (for MIPS.)
Two days ago, a hard cap had been set concerning the attempts to build uClibc in MIPS to emulate it and test the time_t patch’s changes. After today’s changes in the file offset patch and the ones coming up throughout the rest of the week, this just won’t do it. The new time limit for this task is now set to the end of the week. Of course, that does not mean the actual bit-width transitioning efforts are being put on halt.
While reading through the Buildroot docs, a few attempts were made at setting it up through Hashicorp’s Vagrant to avoid having to manage a VM. This was, for the most part, met with failure. Efforts on this front will now switch to building in another VM that I can at least launch (unlike those booted with Vagrant, which oddly seem to hang during guest SSH authentication.)
There’s no news on the currently open PRs.
None at present.
Yesterday’s report set up two new goals for the week; Namely, reviewing the file offset changes in Linux musl and continuing the (considerably larger) uClibc file offset changes. This will likely have to change, to reflect the now (relatively) new need for Buildroot to work and produce a bootable QEMU image with a bootloader, a linux kernel and uclibc-ng as the default c standard implementation, such that testing of the uclibc patch can take place in MIPS. This will have to either replace the goal concerning the musl review for the rest of the week or otherwise take up a few days of it. The change still makes the proposal timeline feasible.