[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: gcc-10-source cross build for xtensa with D enabled fails due to missing texi macro. patch included



No problem.

Both ways (single package or separate package) are fine for me. I did
a single package in the PR draft because it was way easier to do. I
have no idea how to make it into a separate package.

On Sat, 13 Feb 2021 at 18:10, Jonathan McDowell <noodles@earth.li> wrote:
>
> On Sat, Feb 13, 2021 at 10:27:54AM +0000, Witold Baryluk wrote:
> > Have you had time to look at my MR draft I sent before?
>
> This fell off my radar because of the build problems due to not pulling
> in the extra required patches. I've done some cleanup of the package
> build so the extra Debian patches are now applied, which helps matters.
>
> I'm not clear we want to add the D compiler to the default package
> rather than putting it into a separate package?
>
> Also unfortunately at this stage of the release cycle I doubt we can
> make the case for it making bullseye.
>
> > On Fri, 22 Jan 2021 at 16:45, Witold Baryluk <witold.baryluk@gmail.com> wrote:
> > >
> > > A simple (but not-too-simple) test is now included in MR:
> > >
> > > https://salsa.debian.org/electronics-team/toolchains/gcc-xtensa-lx106/-/merge_requests/1/diffs
> > >
> > >
> > > On Fri, 22 Jan 2021 at 15:52, Witold Baryluk <witold.baryluk@gmail.com> wrote:
> > > >
> > > > Yes, for the basic enablement I do have a draft patch (very simple):
> > > >
> > > > https://salsa.debian.org/electronics-team/toolchains/gcc-xtensa-lx106/-/merge_requests/1
> > > >
> > > > By inspection and my own manual tests/use it does work. But a sanity
> > > > test during build process would be indeed a good idea. I will take a look.
> > > >
> > > >
> > > >
> > > >
> > > > On Fri, 22 Jan 2021 at 15:42, Jonathan McDowell <noodles@earth.li> wrote:
> > > > >
> > > > > On Fri, Jan 22, 2021 at 02:22:11PM +0000, Witold Baryluk wrote:
> > > > > > So, it does appear that /usr/src/gcc-10/gcc-10.2.0-dfsg.tar.xz from
> > > > > > gcc-10-source 10.2.1-6 ,
> > > > > > doesn't have these patches applied. And the patches need to be applied
> > > > > > manually after unpacking.
> > > > > >
> > > > > > /usr/src/gcc-10/debian/README.source provides some information, but it
> > > > > > is a bit tricky:
> > > > > >
> > > > > > user@debian:~/xtensa-d/gcc-xtensa-lx106-8$ /usr/src/gcc-10/debian/rules patch
> > > > > > /usr/src/gcc-10/debian/rules:21: debian/rules.patch: No such file or directory
> > > > > > make: *** No rule to make target 'debian/rules.patch'.  Stop.
> > > > > > user@debian:~/xtensa-d/gcc-xtensa-lx106-8
> > > > > >
> > > > > > Try to call make -f without changing directory:
> > > > > >
> > > > > > user@debian:~/xtensa-d/gcc-xtensa-lx106-8$ make -f
> > > > > > /usr/src/gcc-10/debian/rules.patch patch
> > > > > > make: Nothing to be done for 'patch'.
> > > > > > user@debian:~/xtensa-d/gcc-xtensa-lx106-8$
> > > > > >
> > > > > > But, I think this is simply because then debian/rules.patch runs
> > > > > > without proper variables from rules.defs applied, and $(patch_stamp)
> > > > > > is empty string.
> > > > > >
> > > > > > Not sure how to do all the patches, not that xtensa really need most
> > > > > > of them. Just the gdc-texinfo.patch
> > > > >
> > > > > I think the issue here is that I previously attempted to apply the
> > > > > Debian patches, but at some point they stopped being a basic
> > > > > debian_patches.txt file and turned into the current Makefile fragment
> > > > > nightmare. And this wasn't noticed because they're not actually
> > > > > necessary for the lx106 target.
> > > > >
> > > > > Witold, I have no experience with D. I'm happy to enable it for the
> > > > > gcc-xtensa-lx106 package but I'd feel a lot more comfortable doing so if
> > > > > you could provide a basic sanity test to compile like the existing C
> > > > > test in tests/compile-empty-lx106
>
> J.
>
> --
> ] https://www.earth.li/~noodles/ [] I may be cool Beavis, but I can't  [
> ]  PGP/GPG Key @ the.earth.li    []         change the future.         [
> ] via keyserver, web or email.   []                                    [
> ] RSA: 4096/0x94FA372B2DA8B985   []                                    [


Reply to: