problem with bfin-linux-fdpic

[Waldemar Brodkorb]

Hi,
Sam James wrote,

> Waldemar Brodkorb <wbx@openadk.org> writes:
> 
> > Hi Sam,
> > Sam James wrote,
> >
> >> 
> >> 
> >> > On 17 Mar 2025, at 10:15, Waldemar Brodkorb <wbx@openadk.org> wrote:
> >> > 
> >> > Hi Sam,
> >> > Sam James wrote,
> >> 
> >> Hi!
> >> > 
> >> >> Waldemar Brodkorb <wbx@openadk.org> writes:
> >> >> 
> >> >>> Hi,
> >> >>> 
> >> >>> I have an Arcturus Blackfin board I am using for experimenting with
> >> >>> uClibc-ng. It is a bf548 CPU.
> >> >>> Binutils 2.26 with this patch¹ works fine.
> >> >>> Binutils 2.27 does not work anymore.
> >> >>> 
> >> >>> A simple hello world gives me an "Illegal Instruction".
> >> >>> 
> >> >>> I tried to git bisect the problem, but I only get some unrelated
> >> >>> gdb commit as the culprit of the problem.
> >> >>> 
> >> >> 
> >> >> You can pass --disable-gdb when bisecting to not build the other stuff
> >> >> in the repository (see
> >> >> https://github.com/thesamesam/sam-gentoo-scripts/blob/main/gcc/bisect-binutils
> >> >> for a full list of options you can disable for bisecting just binutils).
> >> > 
> >> > Thanks, but I already use --disable-gdb.
> >> > See
> >> > https://cgit.openadk.org/cgi/cgit/openadk.git/tree/toolchain/binutils/Makefile#n56
> >> > 
> >> 
> >> Show me the error you got? May need to disable gdbserver too (possibly other bits as well).
> >
> > I don't get a compile error. The toolchain generation works fine. The problem exist when I am running the
> > generated code. Do you want to see a stacktrace from the target?
> 
> Sorry, I'd clearly misread it - thinking you meant "bisection failed
> because it hit GDB compile errors", my fault.
> 
> Does `git bisect log` show some reasonable good+bad pattern, or was it
> all-good or all-bad? I'd definitely make sure the tools as expected were
> used (note that if using w/ GCC, you'll need to do
> -B/path/to/binutils-pfx for wherever you're bisecting with).

git bisect log
git bisect start
# status: waiting for both good and bad commits
# bad: [cc51fd4ccfdca1559efb0351b498bd8c0f3d1eed] Automatic date update in version.in
git bisect bad cc51fd4ccfdca1559efb0351b498bd8c0f3d1eed
# status: waiting for good commit(s), bad commit known
# good: [2c49145108878e9914173cd9c3aa36ab0cede6b3] Add generated files
git bisect good 2c49145108878e9914173cd9c3aa36ab0cede6b3
# good: [7feec526b7efc8778a5ff536332d4b2a8437dcbb] Add markers for release 2.26
git bisect good 7feec526b7efc8778a5ff536332d4b2a8437dcbb
# good: [7c125e3b10ac6f9222d24c76cdf31a5a9ec8dae0] sim: ppc: avoid use of $< in ordinary rules [PR sim/13834]
git bisect good 7c125e3b10ac6f9222d24c76cdf31a5a9ec8dae0
# good: [35a88fa501ff2e2fe133e7c0f38768ae703a43b7] Fix PR63758 by using the _NSGetEnviron() API on Darwin
git bisect good 35a88fa501ff2e2fe133e7c0f38768ae703a43b7
# good: [2d97cd356e0f0320ecb71cf6a10616ba4618f318] Fix instruction skipping when using software single step in GDBServer
git bisect good 2d97cd356e0f0320ecb71cf6a10616ba4618f318
# good: [ead3d5427a2df5e33316d4ad045510c1d2078c2a] Properly check symbol defined by assignment in linker script
git bisect good ead3d5427a2df5e33316d4ad045510c1d2078c2a
# good: [41d0efca578f03db2787246a4e0c7bf9b085007c] Run gdb.base/sizeof.exp with board having gdb,noinferiorio
git bisect good 41d0efca578f03db2787246a4e0c7bf9b085007c
# bad: [24b368f8532b4775f9fd5bcc7958a84d4218aa46] Fix failures in the GAS testsuite for the ARC architecture.
git bisect bad 24b368f8532b4775f9fd5bcc7958a84d4218aa46
# bad: [7ef8ae7c5f352bb1ef166af0fe6a09a3d3b39f67] Take addend into account when making GOT entries for local symbols.
git bisect bad 7ef8ae7c5f352bb1ef166af0fe6a09a3d3b39f67
# bad: [e58c48b4c81f25ec08174f324990d021aad0e96e] Remove duplicate arch/arm.h include in linux-arm-low.c.
git bisect bad e58c48b4c81f25ec08174f324990d021aad0e96e
# first bad commit: [e58c48b4c81f25ec08174f324990d021aad0e96e] Remove duplicate arch/arm.h include in linux-arm-low.c.

So yes, it shows good+bad pattern.
I always have completely rebuild the toochain between the bisect steps.
 
> I can't help with bfin specifics as I've only ever played with it once, though.

I hope Nick or Alan will dive in.

best regards
 Waldemar