Re: [PATCH 06/12] revert "x86: Also pass -P to $(CPP) when processing i386-opc.tbl"

[Jan Beulich <jbeulich@suse.com>]

On 08.08.2022 14:49, Michael Matz wrote:
> On Fri, 5 Aug 2022, Jan Beulich via Binutils wrote:
>> This reverts commit 384f368958f2a5bb083660e58e5f8a010e6ad429, which
>> broke i386-gen's emitting of diagnostics. As a replacement to address
>> the original issue of newer gcc no longer splicing lines when dropping
>> the line continuation backslashes, switch to using + as the line
>> continuation character, doing the line splicing in i386-gen.
> 
> I will note that all occurrences of the line continuation character are 
> within the various <> templates, never in the opcodes themself.  So their 
> content is all bracketed, and there's no reason why they should even be 
> considered to be single-lined.  So all of this should be possible to be 
> fixed in parse_template() alone, and then i386-opc.tbl wouldn't need any 
> continuation characters at all.

Well - that's possible in principle, yes. But it comes with downsides:
For one, it would require fetching further lines from parse_template(),
when imo it's better to have all fetching (and in particular all
incrementing of "lineno") in a central place (process_i386_opcodes(),
which at present "lineno" also is a local variable of). Then for a
mistakenly omitted closing angle bracket it would lead to consumption
of potentially all remaining lines of the file (quite likely resulting
in a not very helpful diagnostic). Having everything on one (virtual)
line in order to check for the closing angle bracket was actually the
original reason to use line continuation of some form). And finally it
would preclude further use of line continuation for other purposes,
albeit I'll admit doing so may be undesirable for other reasons.

Jan