pdb build fixes

[Jan Beulich]

On 23.12.2022 11:50, Alan Modra via Binutils wrote:
> --- a/ld/emultempl/pe.em
> +++ b/ld/emultempl/pe.em
> @@ -15,6 +15,13 @@ case ${target} in
>      ;;
>  esac
>  
> +case ${target} in
> +  x86_64-*-mingw* | x86_64-*-pe | x86_64-*-pep | x86_64-*-cygwin | \
> +  i[3-7]86-*-mingw32* | i[3-7]86-*-cygwin* | i[3-7]86-*-winnt | i[3-7]86-*-pe)
> +    pdb_support=" ";;
> +  *)
> +esac
> +
>  rm -f e${EMULATION_NAME}.c
>  (echo;echo;echo;echo;echo)>e${EMULATION_NAME}.c # there, now line numbers match ;-)
>  fragment <<EOF
> @@ -66,7 +73,7 @@ fragment <<EOF
>  #include "ldctor.h"
>  #include "ldbuildid.h"
>  #include "coff/internal.h"
> -#include "pdb.h"
> +${pdb_support+#include \"pdb.h\"}

Sadly this doesn't work with older bash (observed with 4.4.23 and 3.2.57), where
the backslashes make it into the resulting ei386pe.c, causing the build to fail.
Consulting documentation I take it that this is a bash bug, but I still think we
need to find a different solution here. I'll try to experiment some, but only
after dealing with another issue that's relevant for 2.40.

Jan