Avoid problems with compilers that wants all variables defined before the code in a function. When Nilsson and wrote the new C-based hash-methods for the Crypto module, I accidently put a variable definition in the middle of the code. GCC didn't complain but some other platforms did.

Instead of learning C better, I developed code that solves the problem... ;)

/ Marcus Agehall (Scanian)

Previous text:

2004-05-05 10:21: Subject: precompile.pike

---

Come again?

/ Peter Bortas (Kein Paket!)

All C-compilers pre C99 and gcc-extensions would complain. I'm not all that enthusiatic about new methods of hiding errors in layers of precompilation.

/ Peter Bortas (Kein Paket!)

Previous text:

2004-05-05 10:31: Subject: precompile.pike

---

Avoid problems with compilers that wants all variables defined before the code in a function. When Nilsson and wrote the new C-based hash-methods for the Crypto module, I accidently put a variable definition in the middle of the code. GCC didn't complain but some other platforms did.

Instead of learning C better, I developed code that solves the problem... ;)

/ Marcus Agehall (Scanian)

How do those compilers want their variables defined then?

/ Marcus Agehall (Scanian)

Previous text:

2004-05-05 10:34: Subject: precompile.pike

---

All C-compilers pre C99 and gcc-extensions would complain. I'm not all that enthusiatic about new methods of hiding errors in layers of precompilation.

/ Peter Bortas (Kein Paket!)

Ah, perhaps I'm very unclear on this point. I wish to do the following:

void f() { for(int x=0;;) ...; }

=>

void f() { int x; for(x=0;;) ...; }

ie, variables shouldn't become global, just moved within the function.

/ Marcus Agehall (Scanian)

Previous text:

2004-05-05 11:18: Subject: precompile.pike

---

In the beginning of each scope. IE, after any "{".

/ Peter Bortas (Kein Paket!)

Trying hard not to.. ;)

The biggest problem IMHO is that things may work just fine on my GCC but might fail on plenty of other platforms. I want to introduce something that prevents this from happening. If you have some better suggestions as how to solve that problem, I'm interested.

/ Marcus Agehall (Scanian)

Previous text:

2004-05-05 11:27: Subject: precompile.pike

---

Learn C.

It would still be likely to introduce bugs.

/ Henrik Grubbström (Lysator)

My problem is that I use C to seldom and use other languages too often (that may look alot like C). This causes some confusion when writing C-code...

/ Marcus Agehall (Scanian)

Previous text:

2004-05-05 12:11: Subject: precompile.pike

---

Learn C.

/ Per Hedbor ()

Is it possible to have GCC warn me of these things? That could be an acceptable solution.

/ Marcus Agehall (Scanian)

Previous text:

2004-05-05 12:29: Subject: precompile.pike

---

There are not that many differences between C89 and C99/gcc in normal day use. You have just learned the most common one.

/ Peter Bortas (Kein Paket!)

That problem can be easily solved with the flags "-pedantic -Werror".

/ Marcus Comstedt (ACROSS) (Hail Ilpalazzo!)

Previous text:

2004-05-05 12:08: Subject: precompile.pike

---

Trying hard not to.. ;)

The biggest problem IMHO is that things may work just fine on my GCC but might fail on plenty of other platforms. I want to introduce something that prevents this from happening. If you have some better suggestions as how to solve that problem, I'm interested.

/ Marcus Agehall (Scanian)

That means that code today that uses for instance

{ .... for (int i=...;;) ... for (float i=...;;) ... }

will break?

/ Mirar

Previous text:

2004-05-05 10:28: Subject: precompile.pike

---

Yes, it would affect cmod-code.

/ Marcus Agehall (Scanian)

You could make it to

{ int x; for(x=0;;) }

instead, which is allowed in most older C compilers too (I think -- at least Pike uses it in a lot of macros).

This will disallow (broken(?)) use of x after the for scope.

/ Mirar

Previous text:

2004-05-05 10:32: Subject: precompile.pike

---

No.

for(int x=0;;)

will become

int x; for(x=0;;)

As far as I know, that should not break things.

/ Marcus Agehall (Scanian)

Yes, then you need to move it to the top of the scope, or create a new scope for x until the end of the current scope:

{

... int x = 10; ...

}

->

{ ... { int x=10; ... } }

/ Mirar

Previous text:

2004-05-05 10:42: Subject: precompile.pike

---

Yes, for those cases that might be possible. I was not aware that such constructions were allowed. But one still needs to move variables defined in the code like:

... int x = 10; ...

should be

int x; ... x=10; ...

since it's not known where x is used after the definition.

/ Marcus Agehall (Scanian)