[Bug symtab/26813] DW_FORM_rnglistx and DW_FORM_loclistx not fully supported

["simark at simark dot ca" <sourceware-bugzilla@sourceware.org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=26813

--- Comment #6 from Simon Marchi <simark at simark dot ca> ---
I'm focusing on rnglists at the moment, but it all probably applies to loclists
too.  I did the same changes as Zoran to make GDB properly handle the rnglistx
form, then hit some more important issues.

First, read_rnglist_index returns a CORE_ADDR: that's wrong, because a
CORE_ADDR is meant to represent an address in the program.  This function
returns an offset in the .debug_rnglists section, the beginning of the range
list.  It should probably be sect_offset.

The .debug_rnglists section in the executable I uploaded looks like this:

[0x00, 0x0B]: header for CU 1
[0x0C, 0x0F]: list of offsets for CU 1 (1 element)
[0x10, 0x78]: range lists data for CU 1

[0x79, 0x84]: header for CU 2
[0x85, 0xB4]: list of offsets for CU 2 (12 elements)
[0xB5, 0xBD7]: range lists data for CU 2

The DW_AT_rnglists_base for CU 2 points to 0x85, that is the beginning of the
list of offsets for that CU.  That's just past the header.

The DW_AT_ranges value for CU 2 has value 0xb (11).  When read_rnglist_index is
called for that index, it reads the header at the beginning of the
.debug_rnglists section.  That's wrong, because the header at the beginning of
the section is the header for CU 1, and we are reading a range list for CU 2. 
If we wanted to read a header, it should be the header at 0x79.  But since the
link from CU 2 to the .debug_rnglists is DW_AT_rnglists_base, and it points
after its header, I don't think we are expected to read the header to do the
index -> range list conversion.  Since the header is not of constant size, we
can't really just read it by going back from the location pointed by
DW_AT_rnglists_base.

We currently use the header to validate that the index is valid, so
unfortunately I think it means we won't be able to do that.  At least, not
unless we do a first pass to slurp the entire .debug_rnglists section to know
where the various contributions are and how many offsets they have.

-- 
You are receiving this mail because:
You are on the CC list for the bug.