🚀 go-pugleaf

In message <m0tMfIe-0000c9C@aladdin.com> L. Peter Deutsch wrote:
> Here is a simple bug that occurs with gcc 2.7.1 and 2.7.2 on many different
> architectures, reported by 3 different users.  I don't have this version of
> gcc readily available to me, so please reply to one of the users in the 'cc'
> list.
>
> A subsequent e-mail states that the compiler produces correctly working
> output if the 'const' is deleted from the cast.


[ Code example of using pointer to const to point to obviously non-const
  deleted ]

> Under 2.7.x, gcc views the pointer dereference in the "if" statement
> as a constant, which is happily evaluates once, outside the loop.

I'm afraid that this is not a bug in gcc; it is a bug in your program
that gcc has uncovered; it is using the fact claim that this is a
const pointer to justify moving the test outside the loop; this is a
good thing.

It seems likely that the only reason for the convoluted `*(const char
*)&x' construct is to prompt compilers towards exactly such optimising
behaviour.  

Perhaps this assumption of const was valid in a previous version of
your code; it clearly no longer is.

> It appears that I will have to put a patch into the next Ghostscript release
> to make it not use 'const' if it is being compiled with these versions of
> gcc.

Why not just replace the construct with "x", and thus fix your
(broken) code?

Am I missing something here?
Rusty.
--
Rusty.Russell@adelaide.maptek.com.au           "Engineer? So you drive trains?"
Lies, damned lies, and out-of-date documentation.        #include <stddisclm.h>