This is the mail archive of the binutils@sources.redhat.com mailing list for the binutils project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Dwarf 2 section flags ?


On Fri, 21 Feb 2003, Alan Modra wrote:

> On Thu, Feb 20, 2003 at 08:56:37AM -0600, Brian Ford wrote:
> > Yes, most bfd targets do this via a strncmp of the section name with
> > ".debug".  In my opinion, that is a gross hack.
>
> I disagree.  In the case of ELF, the ELF ABI explicitly states that
> sections named .debug* contain debugging information.
>
Ok.  This was PE/COFF, though.

> > Unfortunately, since there is currently no equivalent to the .stab
> > directive for DWARF, and there is currently no way to set the
> > debugging flag via the assembly file for sections generated by the
> > compiler, I have to give in and use the section name comparison anyway.
>
> There is no ELF section flag for marking debug sections.  SEC_DEBUGGING
> in bfd is just an internal flag.  How the flag maps to the target format
> must be done in a target dependent way.
>
That is fine, but you deleted the important part of the argument:

In the case I am speaking of, gas is originating these debug sections and
absolutely setting the corresponding section flags to just SEC_READONLY.
Cleary it is wrong for gas to rely on the target bfd back end to detect
via the section name that these are debug sections and fix its inadaquate
section flags setting.  Moreover, the parallel stabs code does set these
flags appropriately.

Shouldn't the internal flag be set consistently when possible so it can
be used by the bfd back end in a target dependent way?

-- 
Brian Ford
Senior Realtime Software Engineer
VITAL - Visual Simulation Systems
FlightSafety International
Phone: 314-551-8460
Fax:   314-551-8444



Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]