Re: [PATCH v2 8/8] Update .debug_names documentation

[Tom Tromey <tom@tromey.com>]

>>>>> "Simon" == Simon Marchi <simark@simark.ca> writes:

>> Since the reader pre-loads all the DIEs I am thinking it probably won't
>> be super expensive to also iterate over them.  Or alternatively we could
>> have the DIE-loading process notice all the spots where an inlined
>> function is seen.

Simon> I was actually wondering if we could avoid pre-loading all the
Simon> DIEs from the start.

This is the approach the cooked indexer takes, so I looked at it for the
full reader.  However the full reader has a few things that make it
harder.

I was mostly looking at this in the context of lazy loading... and there
are even more things that make fully lazy loading hard, like the
"abstract_to_concrete" map and the way that rust_union_quirks currently
works.

Anyway, probably the main issue with lazy DIE loading is that there's
code in the full reader that uses die_info::parent.  This by itself
isn't so bad but it wouldn't work if it's combined with random-access
DIE lookups.

Some of these uses are due to the weird/bad way the full reader computes
names.  Though IIRC there are some uses that are harder to get rid of.

Tom