-
Notifications
You must be signed in to change notification settings - Fork 138
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Builds failing on OSX and Linux with Github Actions CI #320
Comments
I figured it was worth trying to run your Github actions on a fork, seems similar issue - https://github.com/smk762/libwally-core/runs/5672778765?check_suite_focus=true#step:5:549
|
Hi @smk762 Please check Line 151 presumably references an environment variable that is expected to be set to a command (such as If that is the case, you can try setting that variable to the correct value before invoking |
The |
Found a possible solution to this issue. Recently Checking with the upstream version Now
Initially I was under the impression it might be |
Can you share a config.log for this error? |
+1 on this issue. currently trying to build here's my output when running
|
hey @jgriffiths, thanks for looking into this! I just noticed I mis-typed earlier where my logs were from; I was trying to build Anyways, I was still able to produce that output with current
I checked out
hope this is helpful. let me know if there's more info I can help you with collecting |
@jurvis Thanks, I think this error is because its attempting to build a universal binary instead of a singular arch, and that doesn't seem to be supported for adding archive into archives. Can you try limiting the arch to arm64 via |
@jgriffiths yep, it looks like that was it. I was able to build However, when I try and build
here's what the script is doing (abbreviated):
I apologize I don't really have knowledge about building universal binaries and how they're being managed to run on iOS. It looks like everything may be okay on |
@jurvis You may also need to set Aside from allowing M1 builds for wally, my goal is to also have the CI produce wheels that we can release via PyPI so that building from source isn't needed on that platform. That is a question of CI hacking which I am experimenting on in #322. So while its good that you can build wally there, I'll probably consider this closed once we can produce working binaries on the CI, at least for |
hi @jgriffiths I took another stab at building with this script:
I now get the following error:
lmk when that PR is ready to test, I will like to be able to make a contribution to fix downstream for |
@jurivs, at the point at which you get that error, what generated files are in Also I think you do not need |
@jgriffiths here's what I have when doing a hm, I tried using
|
Thanks @jurvis looks like I will need to debug this on an M1 box, I'm arranging access to one. |
This looks like you're mixing secp256k-zkp revisions. The error indicates that libwally is expecting a new version (BlockstreamResearch/secp256k1-zkp@21e2d65 or newer, merged 2022-02-05) because this introduces the "precomputed" files. The dir tree still shows an old revision before this commit. |
@real-or-random Thanks! that looks to be the root cause indeed. @jurvis You need to update the secp submodule in your source tree, #321 is using current master I have one failure to debug/address currently then 321 should be mergable. |
@real-or-random @jgriffiths that was it! can't believe I didn't think of that 🤦♂️ thank you all! |
the script in |
@jurvis Would you able to smoke test the M1 python wheel for me so I can merge #321? Looks like it's going to take a while for me to get access to an M1 box.
|
@jgriffiths sure, I can do it later tonight. will post results here 👍 |
hey @jgriffiths it looks like it worked! Self built wheel
CI-built wheel:
|
@jurvis, thanks very much! I will merge that PR shortly after a little more testing here. |
@jurvis you will need to add the What are you targeting that requires all of the archs listed? |
@jgriffiths I believe |
@jurvis I am no expert on apple silicon/ABIs etc, but I haven't seen anyone explicitly targeting arm64e yet, e.g. python wheels for macos are either I'm guessing its possible that |
@jgriffiths no worries! it may just simply be how I'm managing the ABIs. I'm new to this too myself 😄 I'll continue to poke around and share what I learn when I figure it out. thanks again! |
@jgriffiths I was able to trace and identify the issue. It looks like I encountered my error because
I was able to fix it by just importing |
@jurvis Makes sense, thanks for updating. The current changes add the secp object files to the wally shared library when built, but do not add them to the static library (since its not obvious how to do this in a safe way at this point). gdk builds statically and adds the secp library also, so this solution should continue to work going forward. |
321 is merged, closing. Thanks all for your help resolving this. |
Recently our CI builds at https://github.com/KomodoPlatform/atomicDEX-Desktop/actions have been failing to build libwally.
I suspect it may have something to do with a recent libtools update https://savannah.gnu.org/forum/forum.php?forum_id=10139
related mac errors in log:
Related Linux errors in log:
I've attempted a few things to resolve this without success in our
CI_libwally_test
branch. For example:macos-10.15
instead ofmac latest
in https://github.com/KomodoPlatform/atomicDEX-Desktop/blob/CI_libwally_test/.github/workflows/atomicdex-desktop-ci.ymlOur windows builds are ok, because we used a cached prebuild.
Any hints you might have or insights regarding what might be missing / misconfigured in our workflow file or https://github.com/actions/virtual-environments would be appreciated
The text was updated successfully, but these errors were encountered: