[PATCH v4 2/2] bfd: properly use bfd_get_symbol_leading_char in peXXigen.

[Jan Beulich]

On 01.04.2025 02:31, Jeremy Drake wrote:
> This function returns the leading char to use, so we cannot just assume
> it will always be '_' or '\0'.

Thanks for doing this. I think though that this wants to be patch 1, such
that code you add in what right now is patch 1 is added correctly right
away.

> --- a/bfd/peXXigen.c
> +++ b/bfd/peXXigen.c
> @@ -4405,6 +4405,7 @@ _bfd_XXi_final_link_postscript (bfd * abfd, struct coff_final_link_info *pfinfo)
>    struct coff_link_hash_entry *h1;
>    struct bfd_link_info *info = pfinfo->info;
>    bool result = true;
> +  char buf[20];
> 
>    /* There are a few fields that need to be filled in now while we
>       have symbol table access.
> @@ -4542,10 +4543,10 @@ _bfd_XXi_final_link_postscript (bfd * abfd, struct coff_final_link_info *pfinfo)
>  	}
>      }
> 
> -  h1 = coff_link_hash_lookup (coff_hash_table (info),
> -			      (bfd_get_symbol_leading_char (abfd) != 0
> -			       ? "__tls_used" : "_tls_used"),
> -			      false, false, true);
> +  buf[0] = bfd_get_symbol_leading_char (abfd);
> +  buf[1] = '\0';
> +  strcat(buf, "_tls_used");

This then also wants using in the diagnostics mentioning the symbol name (see
my respective comment on patch 1).

Also, it looks cheaper to me to simply do

  buf[0] = bfd_get_symbol_leading_char (abfd);
  strcpy (buf + !!buf[0], "_tls_used");

? (Further note the blank before the opening parenthesis.)

Finally, instead of naming the buffer "buf" (very unspecific), please consider
using "name" or "symname" or some such.

Jan