On Fri, 2002-11-08 at 09:43, Drew Parsons wrote: > We're not really clear what the problem is though. The bug reporter says > the problem is in imake, but the Xprint imake is supposed to be the same as > in XFree86. As far as I can tell, the package has compiled fine and I > expect it will run fine. > > Could someone running these architectures kindly confirm or deny the > bug, or add more details?
It may well be that the bug doesn't currently manifest itself. However, from the description in the report, it does sound like there is at least a latent problem. If you are using fgetc() and assigning the result to a "char", this is incorrect. The return value is an int. On i386, the only bad effect will be that 0xff bytes are interpreted as EOF, which may never occur in practice; on arm, this bug may cause genuine EOF conditions to be missed, leading to an infinite loop. Compiling with -fsigned-char will make arm behave the same way as i386. However, this will cause a performance loss so using this flag across the board is not desirable. p.
