getline() and the evil widen('\n') call

[rgallaway]

I have a bunch of code that I am porting from Builder 5 to Builder 6. Code that worked fine under BCB5 crashes (throws an exception, really) in the fstream fstream::getline() method. It appears that the stream is trying to access a locale facet that does not exist in the process of executing a call to fstream::widen('\n).

The file in question is a simple ASCII file that has recently been extracted from a blob field of a Paradox db using the SaveToFile() method.

I have tried explicitly setting the locale to classic C locale with no apparent effect.

What is weird that I can read the file with a TRichEdit object just fine, but getline() fails...

What ever happened to nice simple ASCII I/O anyhow? (No offense intended to the several billion people who use non-latin character sets!)

 

[2ffat]

It has to do with how the namespace was defined in BCB5 vs. BCB6. I ran into this with all my BCB5 code and streams. Try using std::getline. Also remember with any ios calls, you also have to define the namespace, too, e.g., std::ios::app.

If you are really using locales, you have to define that namespace. I've been lucky in that I never have had to use anything but the standard namespace.

James P. Cottingham
-----------------------------------------
[sup]I'm number 1,229!
I'm number 1,229![/sup]

 

[butthead]

What ever happened to nice simple ASCII I/O anyhow? (No offense intended to the several billion people who use non-latin character sets!)
[\quote]

have we actually gone so far in our political correctness
to have to apologize for nothing anymore.

butthead

 

[rgallaway]

This problem was resolved by changing from the two parameter form of getline(char * buf, int bufsize) to the three parameter form: getline(char * buff, int bufsize, char terminator). The reason this works is that the two parameter form just calls the three parameter form like this: getline(buf,bufsize,widen('\n')). It was the widen('\n') that was causing the problem anyhow, so I just worked around it.

The downside is that this change involved something like 160 code changes... lots of room for boo boos!