Support the latest newlib, libc++ and GCC (replaces #442) #470
Conversation
|
@ppannuto (and @bradjc & @lschuermann) does this require updating our binary repositories of pre-compiled libraries? |
ppannuto
force-pushed
the
gcc-lib-bump
branch
from
dae33b8
to
3668217
Compare
|
@alevy Yes, I guess so. @bradjc I have never built these toolchains, do you have this infrastructure already set up? We can then add these new artifacts to the mirrorcheck DB which will ensure that we have this replicated everywhere. |
ppannuto
force-pushed
the
gcc-lib-bump
branch
from
3668217
to
2f0dc17
Compare
|
@lschuermann re build infra: that's what the dockerization is for; it should just be a matter of running each of the new That said, I'm building the toolchains currently and can send you built artifacts for the mirror when they finish. |
|
@bradjc @lschuermann Here are updated built artifacts for mirrors: https://www.dropbox.com/scl/fi/26qy6b90048splzzyd42z/libtock-newlib-4.4.0.20231231.zip?rlkey=3wyfyokyoj3i5ukcb9vrotvf5&dl=0 https://www.dropbox.com/scl/fi/ns3ylwnoywgecgjz3vdjy/libtock-libc-14.1.0.zip?rlkey=x79iryk0yq2eme6f381sw4d2p&dl=0 I think this PR should be good to go |
Signed-off-by: Alistair Francis <[email protected]>
Signed-off-by: Alistair Francis <[email protected]>
There was a problem hiding this comment.
Same here for these @ changes. Patch changes make sense.
There was a problem hiding this comment.
The reason is in the title of the commit: show execution of commands that can fail
When running the docker script, it was opaque which command failed if the build failed without this. (i.e., I had to make this change to debug why a build failed).
There was a problem hiding this comment.
That makes sense for debugging but now that it is debugged why would this fail? If this is failing then our docker approach doesn't work.
There was a problem hiding this comment.
I mean, I agree, they shouldn't fail [and don't currently], but if, e.g., something changes when we add the next newlib version I'd rather not build everything twice just to turn on a few prints.
More generally, it aligns with the rest of the build system, i.e., print commands that do something [noting that for the more common / frequent builds with giant commands, we by default shorten (but do not hide) the prints for things like CC foo.c].
@ppannuto Sorry for the delay, added them to my mirror and the mirrorcheck URLs: tock/mirrorcheck@8cdd72e |
|
I guess we can vote, but I feel that these are not commands to build our project but are frozen steps to reproduce a dependency. The changes are also not related to the PR title. |
|
I'm similarly agnostic regarding the |
#442 has an unfortunate chicken-and-egg problem wherein a commit new enough to checkout by hash in the dockerfile will only exist in the main libtock-c repo if the pull request comes from a branch within the libtock-c repo.
Hence this PR, which is based on the external branch, but is now coming from an internal branch. This isn't the most ergonomic or friendly solution, but this is also a rare enough issue [hopefully] that a little occasional human labor seems the path of least resistance.
I added commits to update the repo path, remove the unneeded patch file, and update the hash to check out.
TODO: Technically, the Makefile change to only patch if there's a patchfile is untested as newlib is still building on my laptop, but I doubt there will be any issueHubris will get you every time... ceda51b