It depends, currently the internal structures of radare are mostly raw
single list or arrays of data, and access to it when more than 1000 are
stored the program gets slow.
I have been thinking about the way to reduce this overhead by using caches,
ranges, binary trees and modulus access, but I have not yet decided which
will be the choose.
One of the things I plan to do is to create an universal linked list with
a hashmap and some attributes like string hash within a double linked list,
this way i will be able to access any type of data inside radare on the same
way and be able to index arrays of internal data from the radare shell.
BTW if I have a single data structure for everything I can easily wrap code
and create parsers and mergers for all of them, viewers ...
Some time ago I thought that SQL was the solution for this...but its one
more dependency and I'm not really sure about the performance of this
solution..
For small binaries like hello worlds, store all the dwarf stuff is quite
fast and assumible, but for most of programs dwarf-lines is slow.
In the future the libdwarf will be included inside radare and all these
queries will be performed in realtime like the GNU "addr2line" program
does. And be able to toggle dwarf information file.dwarf=false|true..
Well.. I don't think this changes be prepared for the stable 0.9.3, but
they are on the way and these days I have made the TODO file grow too
much :P so changes and enhacements will come soon.
PD: The same happens with the strings and symbols rsc scripts. Both must
be rewritten in C inside the rabin. With an own ELF/PE parser (GNU one is
buggy and can be fooled), etc.. This way the results will be better and
the program will run faster :)
--pancake
On Sat, Jan 19, 2008 at 09:24:57PM +0100, sha0 sha0 wrote:
> Hi ppl,
>
> I think it will be useful to include dwarf-lines by default, what do you
> think about?
> Why symbols start with sym_, is it possible for simplicity/speed dont
> prepend this?
>
> regards.
>
>
> 2008/1/19, pancake <[EMAIL PROTECTED]>:
> >
> > I've started to move the !hack command as Hack into the core.
> > Currently doesnt works for the debugger, but i'll finish the
> > change soon.
> >
> > Btw thanks to pof to warn me about the missing libexec scripts
> > in Makefile:install.. This make me review the dwarf-lines script
> > to make it work again for the new comment infrastructure.
> >
> > $ cat a.c
> > main()
> > {
> > printf("Fuckmenot!\n");
> > }
> >
> > $ gcc -g a.c
> > $ radare dbg://./a.out
> > (...)
> > [0xB7FA98C0]> .!rsc dwarf-lines $FILE
> > [0xB7FA98C0]> pD 45 @ sym_main
> > 0x0804837C sym_main:
> > ; 2 {
> > 0x0804837C 55 push ebp
> > 0x0804837D 89e5 ebp = esp
> > 0x0804837F 83ec08 esp -= 0x8 ; 8
> > 0x08048382 83e4f0 esp &= 0xf0 ; 240 ; oeax+0xe5
> > 0x08048385 b800000000 eax = 0x0
> > 0x0804838A 83c00f eax += 0xf ; 15 ; oeax+0x4
> > 0x0804838D 83c00f eax += 0xf ; 15 ; oeax+0x4
> > 0x08048390 c1e804 eax >>= 0x4 ; 4 (zerofill)
> > 0x08048393 c1e004 eax <<= 0x4 ; 4
> > 0x08048396 29c4 e
> > Remember that you can move the comments to the right using the
> > asm.cmtmargin
> > and limit the number of comments per offset with asm.cmtlimit.
> > sp -= eax
> > ; 3 printf("Fuckmenot!\n");
> > 0x08048398 83ec0c esp -= 0xc ; 12 ; oeax+0x1
> > 0x0804839B 6874840408 push dword 0x8048474 ; sym__fini+0x24
> > 0x080483A0 e80bffffff ^ call 0x80482B0 ; sym_printf
> > 0x080483A5 83c410 esp += 0x10 ; 16 ; oeax+0x5
> > ; 4 }
> > 0x080483A8 c9 leave ;--
> > 0x080483A9 c3 ret ;--
> > [0xB7FA98C0]>
> >
> > Yeah, thats quite cool for analyzing compilers with dwarf support :)
> >
> > Remember that you can move the comments to the right using the
> > asm.cmtmargin
> > and limit the number of comments per offset with asm.cmtlimit.
> >
> > Have fun! ;D
> >
> >
> > --pancake
> > _______________________________________________
> > radare mailing list
> > [email protected]
> > https://lists.nopcode.org/mailman/listinfo/radare
> >
> -------------- next part --------------
> An HTML attachment was scrubbed...
> URL:
> https://lists.nopcode.org/mailman/private/radare/attachments/20080119/271a3ebb/attachment.htm
>
> _______________________________________________
> radare mailing list
> [email protected]
> https://lists.nopcode.org/mailman/listinfo/radare
>
_______________________________________________
radare mailing list
[email protected]
https://lists.nopcode.org/mailman/listinfo/radare
