Re: dwarflint vs C++

[Roland McGrath <roland at redhat dot com>]

> It checks all EXECINSTR | ALLOC sections for coverage.
> If you point me to crt1.o with debuginfo, I can see if it works, and fix 
> it accordingly, but those that I can get my hands on are debuginfo-less.

One of us is confused (or both!).  I mentioned crt1.o precisely because it
never has debuginfo.  So it contributes to .text but not to any DWARF CU.
(So it is the same as the general case of any statically linked libraries
without debuginfo.)  If you reported all EXECINSTR|ALLOC ranges not covered
by any CU, then you'd already be reporting about the crt1.o regions.

> > (BTW, I think that is a "CU DIE %d" that wants to be a %#x.)
> 
> That's done so because -waranges outputs it as %d.

Oh, I guess that matches what eu-readelf prints in --debug-dump=aranges
and in "Compilation unit at offset %d:", though I was thinking of the DIE
offsets (including the CUs) where it uses %x.  So that is fine unless/until
we decide to make eu-readelf consistently use hex.


Thanks,
Roland