teststute\27_io\basic_filebuf\sgetn\char\1-in.cc

[Karl Lindstrom]

Hello,

I am porting the libstdc++-v3\testsuite to an embedded environment.

I get a few failures in the 27_io\basic_filebuf\sgetn directory.
In 1-in.cc it fails in the void test05() routine at:

   {
     constraint_filebuf fb_01;
     // Need this since BUFSIZ is only guaranteed >= 255 and we want
     // to trigger the same underflow situation everywhere.
     fb_01.pubsetbuf(buffer, 8192);
     fb_01.open(name_01, ios_base::in);
     VERIFY( !fb_01.write_position() );
     strmsz_1 = fb_01.in_avail(); // 8261
     strmsz_2 = fb_01.sgetn(carray1, 10);
     VERIFY( strmsz_2 == 10 );
     strmsz_2 = fb_01.in_avail();
     VERIFY( strmsz_1 > strmsz_2 );
     c1 = fb_01.sgetc();
     VERIFY( c1 == 'b' );
     strmsz_1 = fb_01.in_avail();
     strmsz_2 = fb_01.sgetn(carray2, strmsz_1 + 5);
     VERIFY( strmsz_1 == strmsz_2 - 5 );   // 
<--------------------------------------Fails here.
     c4 = fb_01.sgetc(); // buffer should have underflowed from above.
     VERIFY( c4 == 'h' );
     strmsz_1 = fb_01.in_avail();
     VERIFY( strmsz_1 > 0 );
     strmsz_2 = fb_01.sgetn(carray2, strmsz_1 + 5);
     VERIFY( strmsz_1 == strmsz_2 ); //at the end of the actual file
     VERIFY( !fb_01.write_position() );
     VERIFY( !fb_01.read_position() );
   }

When I look at the code in the debugger:  strmsz_1 and strmsz_2 are  
the same value
at that line.
When I look at the definition of sgetn() it looks like that is  
correct, since asking for
in_avail() + 5 would only return in_avail() characters.

Do I have the latest test, or is there anything I am missing here?

Thanks.

-Karl