llvm dev - Jul 2013 - [LLVMdev] [Patch] WinCOFFObjectWriter: fix for storing pointer to string table in header name field

Hi,

Recently I was hit by an assert in WinCOFFObjectWriter that had forbidden
storing pointer to string table in header name field when the pointer had
more that 6 decimal digits. This limit had been chosen to make
implementation easier (sprintf adds null character at the end) and could be
increased to 7 digits.

My patch is attached. The implementation uses additional buffer on the
stack to make integer to string conversion.

- Paweł Bylica
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20130723/46cda28e/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: WinCOFFObjectWriter_PointerToStingTable.patch
Type: application/octet-stream
Size: 1529 bytes
Desc: not available
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20130723/46cda28e/attachment.obj>

On 23.07.2013 18:29, Paweł Bylica wrote:
> Hi,
>
> Recently I was hit by an assert in WinCOFFObjectWriter that had forbidden
> storing pointer to string table in header name field when the pointer had
> more that 6 decimal digits. This limit had been chosen to make
> implementation easier (sprintf adds null character at the end) and could be
> increased to 7 digits.
I've already implemented this (and also included the undocumented base64 
encoding), see: http://llvm-reviews.chandlerc.com/D667

Maybe someone actually hitting this limit in practice will make someone 
accept at least the first part of D667.

-Nico

Is there a problem if the string is not null terminated?  If not, you can
snprintf it right into place instead of doing sprintf+mempcy.


On Tue, Jul 23, 2013 at 12:29 PM, Paweł Bylica <pawel.bylica at
ibs.org.pl>wrote:
> Hi,
>
> Recently I was hit by an assert in WinCOFFObjectWriter that had forbidden
> storing pointer to string table in header name field when the pointer had
> more that 6 decimal digits. This limit had been chosen to make
> implementation easier (sprintf adds null character at the end) and could be
> increased to 7 digits.
>
> My patch is attached. The implementation uses additional buffer on the
> stack to make integer to string conversion.
>
> - Paweł Bylica
>
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20130723/fba50049/attachment.html>

On 23.07.2013 18:43, Reid Kleckner wrote:
> Is there a problem if the string is not null terminated?  If not, you can
> snprintf it right into place instead of doing sprintf+mempcy.
snprintf always null-terminates (and truncates if there's not enough space).

-Nico

In article <CACs=ty+7zKZU6Ad4jZ5J5Rb0qoMa-bNtO0f+xK_8SYfy3RyFbg at
mail.gmail.com>,
    Reid Kleckner <rnk at google.com> writes:
> Is there a problem if the string is not null terminated?  If not, you can
> snprintf it right into place instead of doing sprintf+mempcy.
Am I the only one who scratches my head and says:

	sprintf?
	memcpy?

Why are we using error-prone C APIs in C++ code?
-- 
"The Direct3D Graphics Pipeline" free book
<http://tinyurl.com/d3d-pipeline>
     The Computer Graphics Museum <http://computergraphicsmuseum.org>
         The Terminals Wiki <http://terminals.classiccmp.org>
  Legalize Adulthood! (my blog) <http://legalizeadulthood.wordpress.com>