--- Log opened Thu Apr 21 00:00:01 2022
00:36 < laanwj> jarolrod: good!
06:41 -!- Netsplit *.net <-> *.split quits: ghost43
07:01 -!- ghost43 [~ghost43@gateway/tor-sasl/ghost43] has joined #bitcoin-core-builds
07:41 -!- sipsorcery [~sipsorcer@2a02:8084:6180:500::12b] has joined #bitcoin-core-builds
10:16 < achow101> hebasto: I'm trying to test out #22555 but I keep getting "ld: error: secp256k1/.libs/libsecp256k1.a(libsecp256k1_la-secp256k1.o) is incompatible with aarch64linux" any idea why that is?
10:16 < gribble> https://github.com/bitcoin/bitcoin/issues/22555 | build: Fix `make apk` for Android w/ non-default SOURCES_PATH in depends by hebasto · Pull Request #22555 · bitcoin/bitcoin · GitHub
10:16 < achow101> it seems to happen on master too
10:33 < hebasto> achow101: let me look into it
10:34 < sipa> in a clean build tree?
10:34 < achow101> yes
10:35 < achow101> could ccache cause this?
10:35 < sipa> It shouldn't, but if you want to make sure, `ccache -C` clears the cache.
10:37 < achow101> yeah, no effect
10:39 < sipa> Is https://github.com/bitcoin-core/secp256k1/pull/1104 in?
10:39 < sipa> No, it can't be - the current subtree in bitcoin core doesn't have it.
10:41 -!- sipsorcery [~sipsorcer@2a02:8084:6180:500::12b] has quit [Ping timeout: 248 seconds]
10:44 < sipa> I don't think that'd fix it, but who knows.
10:48 < achow101> nope, no effect
10:50 -!- sipsorcery [~sipsorcer@2a02:8084:6180:500::12b] has joined #bitcoin-core-builds
11:06 < laanwj> achow101: it looks like that file is compiled with the wrong C compiler, is it using the cross-compiler, or the wrong arguments? would make sense to build with V=1
11:07 < laanwj> it's the only C file compiled as part of bitcoin core so if the C compiler is misconfigured, it'd turn up there
11:11 < achow101> laanwj: seems plausible. It looks like everything else uses clang provided by the ndk whereas this file is being compiled with gcc (presumably system gcc?)
11:14 < laanwj> achow101: maybe one of the build changes stopped overriding CC, though then it's strange that the other cross builds don't fail
11:15 < laanwj> I would assume it's a silent merge conflict 
11:16 < laanwj> we discussed compiling secp256k1 with the C++ compiler at some point but it was deemed too risky
11:16 < hebasto> https://github.com/bitcoin-core/secp256k1/pull/988 ?
11:17 < fanquake> if 22555 doesn't compile why isn't the CI broken?
11:17 < achow101> do we have ci for android builds?
11:17 < fanquake> https://github.com/bitcoin/bitcoin/pull/22555/checks?check_run_id=6113931756
11:18 < fanquake> I assume that's what that is
11:18 < laanwj> maybe the CI would fail now (with current master)
11:18 < achow101> oh
11:18 < achow101> It was just rebased, so I don't think it's an out of date branch thn
11:18 < hebasto> laanwj: it is just freshed rebased
11:18 < laanwj> ok
11:19 < laanwj> no idea then
11:39 < achow101> cross compiling 22555 for windows seems to just not be working at all for me too.
11:43 < achow101> and not on master either
11:44 < hebasto> what error for windows build?
11:44 < achow101> just a bunch of undefined references for basically everytning in secp
11:46 < laanwj> how are you cross-compiling?
11:47 < laanwj> i didn't try master, but guix-built  #24866 for all platforms today and there were no errors
11:47 < gribble> https://github.com/bitcoin/bitcoin/issues/24866 | build: No longer need to hack the `PATH` variable in `config.site` by hebasto · Pull Request #24866 · bitcoin/bitcoin · GitHub
11:48 < achow101> the usual depends way
11:48 < laanwj> with a clean tree?
11:48 < achow101> yea
11:50 < laanwj> so it sounds like a similar issue as before, with the wrong C compiler being used for secp256k1
11:52 < achow101> yes, if I look at src/secp256k1/Makefile, CC is just gcc
11:52 < achow101> but if I look at the same file in the guix build distsrc, it's x86_64-w64-mingw32-gcc
11:52 < achow101> (guix build of 23.0rc5)
11:53 < achow101> I'm trying to bisect to find where this happened
11:58 < laanwj> good idea
12:02 < hebasto> hi
12:38 < achow101> I figured it out. I forgot to put pwd in my CONFIG_SITE
12:38 < achow101> I thought it would be able to figure out the relative path, but I guess not
12:39 < hebasto> yes, it requires absolute paths only
12:39 < achow101> a warning about that would be nice...
12:49 < laanwj> it works for the first level of configure, but not the nested one
12:49 < laanwj> which, i guess, is invoked from a different cwd
12:51 < laanwj> ideally configure would fail if provided a CONFIG_SITE that doesn't exist, not ignore it and continue as if nothing happens, that always results in the most annoying and hard to debug issues
12:57 < laanwj> CONFIG_SITE handling is entirely internal to autoconf so i don't think it's easy to change
13:31 -!- ghost43 [~ghost43@gateway/tor-sasl/ghost43] has quit [Quit: Leaving]
13:32 -!- ghost43 [~ghost43@gateway/tor-sasl/ghost43] has joined #bitcoin-core-builds
15:46 -!- sipsorcery [~sipsorcer@2a02:8084:6180:500::12b] has quit [Ping timeout: 240 seconds]
23:18 -!- sipsorcery [~sipsorcer@37.228.225.67] has joined #bitcoin-core-builds
--- Log closed Fri Apr 22 00:00:02 2022