| 19 Apr 2025 |
|  alexfmpe changed their profile picture. | 23:55:44 |
| 20 Apr 2025 |
 Colin | anyone given thought about how to get blueprint-compiler to be cross friendly? | 02:54:51 |
| Colin | it's used by a lot of gtk4 apps. runs at compile time and wants to load the GI typelibs used by the application. not sure how to wrangle that except by building an extra copy of every runtime dependency, but targeting the build machine, so that blueprint-compiler can import it | 02:57:11 |
 dramforever | one possible way is to use an emulator of the host platform | 03:27:18 |
| dramforever | try grepping stdenv.hostPlatform.emulator nix nixpkgs for examples | 03:27:53 |
| dramforever | * try grepping stdenv.hostPlatform.emulator in nixpkgs for examples | 03:27:56 |
| dramforever | ... you know you have nix brain when you typo "in" into "nix" | 03:28:15 |
|  spikespaz set a profile picture. | 04:26:18 |
| 21 Apr 2025 |
 Tristan Ross | Trying to fix pkgsLLVM.linux to work correctly, got this odd issue:
linux> make -f /build/linux-6.6.87/tools/build/Makefile.build dir=. obj=fixdep
linux> /nix/store/rpxlridr3yn6mbmd34r0bya9qv91b79s-gcc-wrapper-14.2.1.20250322/bin/cc -o /build/linux-6.6.87/build/tools/bpf/resolve_btfids/fixdep /build/linux-6.6.87/build/tools/bpf/resolve_btfids/fixdep-in.o
linux> mkdir -p /build/linux-6.6.87/build/tools/bpf/resolve_btfids//libsubcmd
linux> make -C /build/linux-6.6.87/tools/lib/subcmd/ OUTPUT=/build/linux-6.6.87/build/tools/bpf/resolve_btfids/libsubcmd/ \
linux> DESTDIR=/build/linux-6.6.87/build/tools/bpf/resolve_btfids/libsubcmd/ AR="/nix/store/0w9bd0ljiyj2crmywkm0mvm4nwsswgb0-binutils-wrapper-2.44/bin/ar" CC="/nix/store/rpxlridr3yn6mbmd34r0bya9qv91b79s-gcc-wrapper-14.2.1.20250322/bin/cc" LD="/nix/store/0w9bd0ljiyj2crmywkm0mvm4nwsswgb0-binutils-wrapper-2.44/bin/ld" ARCH="arm64" CROSS_COMPILE="" EXTRA_CFLAGS="" prefix= subdir= \
linux> /build/linux-6.6.87/build/tools/bpf/resolve_btfids/libsubcmd/libsubcmd.a install_headers
linux> exec-cmd.c:2:10: fatal error: linux/compiler.h: No such file or directory
linux> 2 | #include <linux/compiler.h>
linux> | ^~~~~~~~~~~~~~~~~~
linux> compilation terminated.
linux> make[6]: *** [/build/linux-6.6.87/tools/build/Makefile.build:98: /build/linux-6.6.87/build/tools/bpf/resolve_btfids/libsubcmd/exec-cmd.o] Error 1
linux> make[5]: *** [Makefile:80: /build/linux-6.6.87/build/tools/bpf/resolve_btfids/libsubcmd/libsubcmd-in.o] Error 2
linux> make[4]: *** [Makefile:58: /build/linux-6.6.87/build/tools/bpf/resolve_btfids//libsubcmd/libsubcmd.a] Error 2
linux> make[3]: *** [Makefile:76: bpf/resolve_btfids] Error 2
linux> make[2]: *** [/build/linux-6.6.87/Makefile:1362: tools/bpf/resolve_btfids] Error 2
linux> make[1]: *** [/build/linux-6.6.87/Makefile:234: __sub-make] Error 2
linux> make[1]: Leaving directory '/build/linux-6.6.87/build'
linux> make: *** [Makefile:234: __sub-make] Error 2
| 00:12:07 |
| dramforever | what's the command for the actual build of exec-cmd.o? | 00:16:55 |
| Tristan Ross | Idk, I don't see it | 00:18:09 |
| dramforever | V=1 to makeflags probably | 00:22:58 |
| dramforever | * V=1 to makeFlags probably | 00:23:03 |
| Tristan Ross | That's already set | 00:23:14 |
| dramforever | huh, really? no exec-cmd.o at all in log? | 00:23:34 |
| Tristan Ross | I added CFLAGS=-I$(readlink -e include) and get even worse errors | 00:23:36 |
| Tristan Ross | None | 00:23:40 |
| Tristan Ross | Because it's a make inside a make | 00:23:45 |
| dramforever | i don't see why that matters | 00:24:27 |
| Tristan Ross | It's not passing through the V | 00:24:59 |
| Tristan Ross | So when make is calling make, the toplevel make is verbose | 00:25:09 |
| Tristan Ross | But the child make isn't | 00:25:13 |
| dramforever | i'm pretty sure that is not how makefile works | 00:25:25 |
| Tristan Ross | Well, one of the lines is literally this:
linux> make -C /build/linux-6.6.87/tools/lib/subcmd/ OUTPUT=/build/linux-6.6.87/build/tools/bpf/resolve_btfids/libsubcmd/ \
linux> DESTDIR=/build/linux-6.6.87/build/tools/bpf/resolve_btfids/libsubcmd/ AR="/nix/store/0w9bd0ljiyj2crmywkm0mvm4nwsswgb0-binutils-wrapper-2.44/bin/ar" CC="/nix/store/rpxlridr3yn6mbmd34r0bya9qv91b79s-gcc-wrapper-14.2.1.20250322/bin/cc" LD="/nix/store/0w9bd0ljiyj2crmywkm0mvm4nwsswgb0-binutils-wrapper-2.44/bin/ld" ARCH="arm64" CROSS_COMPILE="" EXTRA_CFLAGS="" prefix= subdir= \
linux> /build/linux-6.6.87/build/tools/bpf/resolve_btfids/libsubcmd/libsubcmd.a install_headers
| 00:26:03 |
| Tristan Ross | And I don't see a V=1 | 00:26:16 |
| Tristan Ross | The source is this: https://github.com/torvalds/linux/tree/v6.6/tools/bpf/resolve_btfids | 00:26:35 |
| dramforever | on one hand, V=1 is definitely passed down | 00:26:58 |
| dramforever | the way it works in linux is in the top level Makefile it actually goes, if V set in command line, set KBUILD_VERBOSE to that | 00:27:22 |
| Tristan Ross | Huh, even when it's building the host stuff? | 00:27:38 |
| dramforever | and then it sends these down in environment variablesexport quiet Q KBUILD_VERBOSE | 00:27:43 |
