Hi Paul, On 2026-11-03 00:17, Paul Boddie wrote:
On 2026-11-02 18:31, Marcus Haehnel wrote:
I wonder a bit why your l4re-core/libgcc_s directory comes from, which you build with the make command you cite above. There is no such directory and never has been. Is this maybe something you have locally? Or did you mean S=pkg/l4re-core/libgcc?
No, this is a real directory! It was installed by ham when I ran these commands:
ham init -u https://github.com/kernkonzept/manifest.git ham sync
Sorry, that was my fault for not checking the latest version of the source tree. You are of course right. The directory is there. See also my latest mail to the list. I retracted the mail you replied to from the list due to this mistake by me, since it was still in moderation. But forgot that you got it anyways ;)
Usually things like libgcc come in through the compiler-rt .pc file (see pc directory in the build dir) and depend on your config (i.e. if you use the toolchain libgcc or the ones from l4re-core). The compiler-rt should be pulled in by other packages, usually there is no need to directly depend on it.
Did you make an `make oldconfig` after the update, to make sure the configuration is up to date? Usually we default to using our supplied libgcc nowadays but the build must be correctly configured for this to work.
I probably should have noted this, but I was populating the build tree from scratch, also copying my own packages into the pkg directory. I don't really know how to use ham beyond setting up the packages in the manifest, so I couldn't say where I might have gone wrong.
[Errors]
So you did `make oldconfig` and / or a global make from the root directory? See also the other mail. Just doing `make S=pkg/foo/bar` will not build dependencies. `make pkg/foo/bar` will and should trigger a build of the libgcc_s. To check if the rules have been regenerated correctly by `make oldconfig` have a look into .Package.deps in the build dir. All programs build shared, and all libraries that also have a shared variant (.so) should depend on `stdlibs-sh` which depends on `libgcc_s`. I checked on my build tree and this is the case for me.
Sadly this is not easy due to the parallel building inside the packages. But I have thought about this before and might put it higher on my agenda that your main make that triggered the build reads a result file for failed packages. But it is not straight forward to do this.
I appreciate that it is difficult, certainly. Make isn't the easiest tool to work with.
Please let me know if anything of the above (i.e. the `oldconfig`) helped, and if not please let me know how your L4Re is configured, in particular what the settings of the `CONFIG_COMPILER_RT_*` options are.
From mybuild/.config:
CONFIG_COMPILER_RT_CHOICE_AUTO=y CONFIG_COMPILER_RT_USE_LIBGCC=y
This looks correct. Please check .Package.deps. Best, - Marcus
Thanks for the response!
Paul _______________________________________________ l4-hackers mailing list -- l4-hackers@os.inf.tu-dresden.de To unsubscribe send an email to l4-hackers-leave@os.inf.tu-dresden.de