[PATCH v5 03/12] BFD: Correct archive format determination for non-default link emulations

[Hans-Peter Nilsson]

> Date: Wed, 28 Jan 2026 10:55:14 +0000
> From: "Maciej W. Rozycki" <macro@orcam.me.uk>

Trying to fast-forward through this patch-set which looks
acceptable regarding CRIS changes, but this text...

> From: Alan Modra <amodra@gmail.com>
> 
> Fix an issue with archive format determination coming from commit 
> b228303dd14e ("PR binutils/13278: --plugin doesn't work on archive"), 
> <https://inbox.sourceware.org/binutils/20111011064345.GM25970@bubble.grove.modra.org/>, 
> causing the default link emulation to be used despite archive members 
> being in another object format and consequently triggering segmentation 
> faults in `cris-aout' LD with the default (`crisaout') emulation where 
> archive members are in the ELF format.
> 
> Address the issue by always checking the format of the first member of 
> an archive even if no default target has been specified, except when the 
> target being checked against is be the plugin target.  Verified by hand 

(nit: "is be the" sounds like wrong grammar)

> and the `powerpc64le-linux-gnu' target to cause no issue with the recipe 
> in PR binutils/13278 and removing failures from segmentation faults with 
> the new tests:
> 
> cris-aout  -FAIL: CRIS regular archive link (implicit emulation)
> cris-aout  -FAIL: CRIS thin archive link (implicit emulation)
> 
> Co-Authored-By: Maciej W. Rozycki <macro@redhat.com>
> ---
> No change from v4 (05/14), 
> <https://inbox.sourceware.org/binutils/alpine.DEB.2.21.2601072208110.30566@angie.orcam.me.uk/>:
> 
> Changes from v3 (11/18), 
> <https://inbox.sourceware.org/binutils/alpine.DEB.2.21.2511192246460.57987@angie.orcam.me.uk/>:
> 
> - Retain the check for a defaulted target and instead also enable the
>   first archive member format check for linker input.
> 
> - Request that no plugin be used for the first archive member format 
>   check.
> 
> New change in v3.

...does not explain to me why...

> ---
>  bfd/archive.c                    |    5 ++++-
>  ld/testsuite/ld-cris/archive.exp |   21 +++++++++++++++++++++
>  2 files changed, 25 insertions(+), 1 deletion(-)
> 
> binutils-bfd-archive-format-check-fix.diff
> Index: binutils-gdb/bfd/archive.c
> ===================================================================
> --- binutils-gdb.orig/bfd/archive.c
> +++ binutils-gdb/bfd/archive.c
> @@ -928,7 +928,8 @@ bfd_generic_archive_p (bfd *abfd)
>        return NULL;
>      }
>  
> -  if (abfd->target_defaulted && bfd_has_map (abfd))
> +  if ((abfd->target_defaulted || abfd->is_linker_input)
> +      && bfd_has_map (abfd))
>      {
>        bfd *first;
>        unsigned int save;
> @@ -951,6 +952,8 @@ bfd_generic_archive_p (bfd *abfd)
>        if (first != NULL)
>  	{
>  	  first->target_defaulted = false;
> +	  if (abfd->is_linker_input)
> +	    first->plugin_format = bfd_plugin_no;
>  	  if (!bfd_check_format (first, bfd_object)
>  	      || first->xvec != abfd->xvec)
>  	    bfd_set_error (bfd_error_wrong_object_format);
> Index: binutils-gdb/ld/testsuite/ld-cris/archive.exp
> ===================================================================
> --- binutils-gdb.orig/ld/testsuite/ld-cris/archive.exp
> +++ binutils-gdb/ld/testsuite/ld-cris/archive.exp
> @@ -68,6 +68,14 @@ run_ld_link_tests [list \
>  	 {nm "" ../ld-archive/nx.nd}} \
>  	"abcx" \
>      ] \
> +    [list "CRIS regular archive link (implicit emulation)" \
> +	"-e ff" "tmpdir/ab.a --verbose --print-map" \
> +	"$asemul" {../ld-archive/abc.s} \
> +	{{ld ../ld-archive/abc.vd} \
> +	 {nm "" ../ld-archive/abc.nd} \
> +	 {nm "" ../ld-archive/nx.nd}} \
> +	"abci" \
> +    ] \
>  ]
>  if { ![istarget cris-*-*aout*] } {
>      setup_xfail "binutils/33485" "*-*-*"
> @@ -82,3 +90,16 @@ run_ld_link_tests [list \
>  	"abtcx" \
>      ] \
>  ]
> +if { ![istarget cris-*-*aout*] } {
> +    setup_xfail "binutils/33485" "*-*-*"
> +}

...the following test needs to be xfailed for non-aout CRIS
configurations (FAOD I'm looking for a logical explanation
why they'd fail, i.e. not just "because they're failing").
Probably just a line or two is needed.

> +run_ld_link_tests [list \
> +    [list "CRIS thin archive link (implicit emulation)" \
> +	"-e ff" "tmpdir/abt.a --verbose --print-map" \
> +	"$asemul" {../ld-archive/abc.s} \
> +	{{ld ../ld-archive/abtc.vd} \
> +	 {nm "" ../ld-archive/abc.nd} \
> +	 {nm "" ../ld-archive/nx.nd}} \
> +	"abtci" \
> +    ] \
> +]
> 

Otherwise ok.

brgds, H-P