Re: [MiNT] g++ bug or is my build enviroment not sane?

[George Nakos <ggn@hol.gr>]

Hello Vincent,

Tuesday, October 27, 2009, 12:35:42 AM, you wrote:

> ggn (HOL) wrote:
>>   resulted  in  only the first console input (read by cin) to be read, while the
>>   rest  were  read  as  blank  (presumably,  you  don't  get  a  chance to enter
>>   anything).
> ...
>>   aranym.  Both  machines  have Mikro's 4.4.1 gcc port and 2.19.1 binutils. Also

> This is the long lasting cout bug. As I said many times, it occurred after 
> outputting an int to cout: the output stream was flagged as error, 
> preventing anything else to be printed after that.

Well,  as  you see from the source (which fails on my config), no printing a int
using cout is performed.

> I advertised last week that I fixed that bug in binutils 2.19.1 20091017.

> I confirm your program works well with GCC 4.4.2 20091017 and binutils 2.20 
> 20091024. I compiled your program and ran it successfully on simple ST with 
> plain TOS. I must admit the resuting executable optimized and stripped is 
> 396698 bytes big, it is quite a shame.

> Upgrading the binutils is not enough, because the C++ libraries provided 
> with GCC are compiled with the old binutils, so they are wrong. You must 
> upgrade GCC and the binutils. There are currently no fixed binaries 
> available for running on MiNT.

So, the solution is downloadin binutils 2.20, use the patch from your site, then
recompiling mintlib?

> Note that you reduced the code too much, you read only values, but nothing 
> is displayed. I had to add some lines to display the values in order to test 
>   your program.

> No mystery here.

As  I  wrote  above,  a  cout  to  print  an integer is not needed for it to not
function (not even the original source does something like that).


-- 
Best regards,
 George                            mailto:ggn@hol.gr