Re: [RFC PATCH] Implement native TLS support on Windows

[Tom Tromey <tom@tromey.com>]

>>>>> "Hannes" == Hannes Domani <ssbssa@yahoo.de> writes:

Hannes> I've used the fetch_tls_load_module_address gdbarch method to get the
Hannes> address of _tls_index of the OBJFILE, which is then forwarded as LM_ADDR
Hannes> to windows_get_thread_local_address, but I'm not really sure if this is
Hannes> allowed.

I think it is fine, or at most would need a small update to the comment
string explaining that the values are opaque and intended to be
target-specific.

Hannes> +/* Get the address of variable "_tls_index" of OBJFILE.  */
Hannes> +
Hannes> +static CORE_ADDR
Hannes> +windows_tls_index_address (struct objfile *objfile)
Hannes> +{
Hannes> +  bound_minimal_symbol minsym
Hannes> +    = lookup_minimal_symbol_linkage("_tls_index", objfile, false);

Missing space.

Hannes> +  gdb_byte buf[8];
Hannes> +  if (target_read_memory (lm_addr, buf, 4))
Hannes> +    throw_error (TLS_GENERIC_ERROR, _("Cannot read _tls_index"));
Hannes> +
Hannes> +  enum bfd_endian byte_order = gdbarch_byte_order (gdbarch);
Hannes> +  uint32_t tls_index = extract_unsigned_integer (buf, 4, byte_order);
Hannes> +  CORE_ADDR tlb, tls_ptr, slot_ptr;
Hannes> +  if (target_get_tib_address (ptid, &tlb)
Hannes> +      && !target_read_memory (tlb + tls_offset, buf, ptr_bytes)
Hannes> +      && (tls_ptr = extract_unsigned_integer (buf, ptr_bytes,
Hannes> +					      byte_order)) != 0
Hannes> +      && !target_read_memory (tls_ptr + tls_index * ptr_bytes,
Hannes> +			      buf, ptr_bytes)
Hannes> +      && (slot_ptr = extract_unsigned_integer (buf, ptr_bytes,
Hannes> +					       byte_order)) != 0)

Instead of multiple assignments in an 'if' I think it'd be better to
have separate 'if's and early returns.

Tom