Re: [PATCH 2/2] gdb: use std::vector<> to hold on blocks in struct blockvector

[Simon Marchi <simark@simark.ca>]

On 10/14/25 11:00 AM, Andrew Burgess wrote:
>> @@ -809,6 +810,41 @@ make_blockranges (struct objfile *objfile,
>>    return blr;
>>  }
>>  
>> +static bool
>> +block_ordering_predicate (struct block *b1, struct block *b2)
> 
> Functions should have a comment above them.
> 
>> +{
>> +  CORE_ADDR start1 = b1->start ();
>> +  CORE_ADDR start2 = b2->start ();
>> +
>> +  if (start1 != start2)
>> +    return start1 < start2;
>> +  return (b2->end () < b1->end ());
>> +}
> 
> I notice that we use a slightly different ordering function in
> buildsym.c, one that doesn't consider the end address.  Probably doesn't
> matter, but maybe it's at least worth mentioning in a comment or the
> commit message, that way, if someone is looking at the code in the
> future they can understand why these things are different.  Even if the
> reason is just "we have no motivating example for why these should be
> the same right now".

buildsym sorts them (locally, in its singly-linked list) in the inverse
order, but then fills the blockvector backwards.  So the blocks end up
in the right order in the blockvector.  I think this is an artefact from
the past and could be simplified.  The predicate for sorting blocks in a
blockvector should probably be written just once (in block.h) and used
by whoever needs it.

Simon