This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: SS 20020521 ostream.tcc breaks ostream output


tmwg-gcc@inxservices.com wrote:

>On Thu, May 23, 2002 at 01:17:13PM +0200, Paolo Carlini wrote:
>  
>
>>For the time being, you may workaround the problem by always checking 
>>/explicitly/ that the const char* passed to ostream::operator<<(const 
>>char*) is not null, doing nothing in that case, i.e., not calling 
>>operator<< at all.
>>(f..i, for the lines "temp_file << ident_buffer << '\n';" and "temp_file 
>><< buffer << '\n';")
>>
>>Ciao, Paolo.
>>    
>>
>   How can either ident_buffer or buffer be null in this case, since
>they are both arrays of chars on the stack? I presume you mean
>buffer[0] == 0?
>
Indeed.

>   Do you realise how much code has to handle empty lines, that is lines
>that are only a new line? And all this will now either be audited or
>break? Is that what is being said?
>   Because that is the symptom. There is an empty line where the output
>stream gets the fail bit.
>
I said "for the time being" and "workaround", right?

This implies that, depending on the opinion of Benjamin Kosnik, v3 
maintainer and author of the commit which changed the behaviour of such 
kind of code, part of the previous behaviour will be restored, that is 
the bad bit /not/ set in such situations.

Ciao, Paolo.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]