Hi, Mike

    What you report here is indeed *grave*

    I am about to package version 3.2 per another user's request
(otherwise, I would have requested this package's removal from the
Archive in light of this)

Mike Hommey wrote:
> severity 488467 grave
> retitle 488467 libtrash is dangerously broken on amd64
> tag 488467 patch
> thanks
>
> On Sun, Jun 29, 2008 at 09:25:13AM +0200, Mike Hommey wrote:
>   
>> On Sun, Jun 29, 2008 at 09:10:00AM +0200, Mike Hommey wrote:
>>     
>> Here is the culprit:
>> real_fopen = dlvsym(RTLD_NEXT, "fopen", "GLIBC_2.1");
>>
>> There is no such version of the symbol:
>> $ objdump -T /lib/libc.so.6 | grep ' fopen$'
>> 0000000000064840 g    DF .text  000000000000000a  GLIBC_2.2.5 fopen
>>     
>
> It's even worse than that. Anything using other f* functions (freopen,
> fopen64, etc.) will get a wrong return value because of bad declarations
> for the function pointers.
>   
This is very serious
> The attached patch should fix both issues.
Looks very nice, thanks.
> It doesn't fix more of the brain damage in the code source itself (dlsym()ing 
> at every call is
> stupid,
indeed
>  having only one function to do the whole thing is bad design,
>   
completely agreed
> and macros that are used only once don't help making the code readable)
>   
> Although this patch makes things worse, I strongly advise to audit the
> code for other similar "jokes". (Or, even better, a complete rewrite)
>   
I seem to remember that version 3 was mostly a rewrite ... we'll be able
to tell very soon.


Would you mind sponsoring an upload of version 3.2 ?
I think I can have it ready pretty soon.



I stopped using libtrash quite a while ago, and have only kept for some
users' sake (I know what it feels when one maintainer decides to drop
that package you have come to depend on). I was considering requesting
its removal seeing that nobody seemed to use it anymore.

Recently, an user has contacted me requesting a newer version and
reporting some very active users, so I
 decided to package the latest version.... and then you came with this,
definitively making it necessary.



I will send you an e-mail as soon as I have the new package ready.

Thank you for your interest and work, Mike.


    J.L.




-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Reply via email to