This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [PATCH 0/3] remove-symbol-file
- From: Pedro Alves <palves at redhat dot com>
- To: "Blanc, Nicolas" <nicolas dot blanc at intel dot com>
- Cc: Tom Tromey <tromey at redhat dot com>, "gdb-patches at sourceware dot org" <gdb-patches at sourceware dot org>
- Date: Fri, 26 Apr 2013 18:33:06 +0100
- Subject: Re: [PATCH 0/3] remove-symbol-file
- References: <1366098721-18302-1-git-send-email-nicolas dot blanc at intel dot com> <8761zdow5j dot fsf at fleche dot redhat dot com> <388084C8C1E6A64FA36AD1D656E4856619DEEE98 at IRSMSX102 dot ger dot corp dot intel dot com> <51781A68 dot 4010004 at redhat dot com> <388084C8C1E6A64FA36AD1D656E4856619DEF279 at IRSMSX102 dot ger dot corp dot intel dot com>
On 04/25/2013 09:38 AM, Blanc, Nicolas wrote:
>
>
>>> session. In this context I think the text address is the most appropriate way to remove a file because:
>>> 1) the user knows exactly where the .text section was loaded,
>>
>> So how do you handle the case of there being no .text section at all?
>
> It's a requirement of the add-symbol-file command [1]. The existing implementation of add-symbol-file assumes
> that the *mandatory* load-address argument (terminology from GDB's manual) is the address of the text section.
> See the implementation of add-symbol-file:
>
> if (argcnt == 1)
> {
> /* The second argument is always the text address at which
> to load the program. */
> sect_opts[section_index].name = ".text";
>
> The second argument is not optional. The add-symbol-file command returns an error if no address is specified.
As I mentioned before, in the add case, you can just pass in a
random ADDR, I think gdb copes:
(gdb) add-symbol-file ~/data.o 0x111 -s .data 0x2000
add symbol table from file "/home/pedro/data.o" at
.text_addr = 0x111
.data_addr = 0x2000
(y or n) y
Reading symbols from /home/pedro/data.o...warning: section .text not found in /home/pedro/data.o
>>> Note that currently in Option 4 below ADDR is in fact
>>> "objf->addr_low", but the command could be more generous by searching
>>> first which file corresponds to ADDR and then removing it. This would be more flexible and an alternative to Option 3, for instance.
>> You lost me here.
>
> It's an idea that you and Tom gave me. Given an arbitrary address, remove-symbol-file could figure out the file that
> corresponds to this address and remove it. This could make the command more flexible.
Hmm, how is that different from what the command is currently
doing? /me looks
> But thinking twice I think that the current
> implementation of remove-symbol-file does the right thing by identifying the file to remove using the address from the
> add-symbol-file command. The user knows what address he passed to add-symbol-file.
Ah, so the current implementation stores the ADDR the user specified
with add-symbol-file. Why didn't you say so? ;-)
Okay, I guess that's good enough. We can add support for "-s" too
to remove- if ever necessary.
>>> 1) remove-symbol-file FILE
>>> 2) remove-symbol-file FILE ADDR
>>> 3) remove-symbol-file -s .data DATA_ADDR
>>> 4) remove-symbol-file ADDR
--
Pedro Alves