Re: [Tinycc-devel] Small curiosity, re operator precedence.

[Paulo Henrique Torrens]

Well, let's see...

1. while((*--o=*--i)!=92)
2. while(92!=*--o=*--i)

It's quite important that assignment have more priority, for simple expressions as "bool x = a != b" to work. By the standards, (...).

Forget what I was going to say, I just think I figured it out.
Yeah, != has bigger precedence, but look what you did: it is on the left side of the _expression_, so it wouldn't be a valid syntax that way. :)
Probably GCC just skips this syntax, and I'm not sure if the second option (seeing the *--o = *--i as the right side) is invalid in an ANSI application, but TCC's accepting it would be like an "extension", ain't it? I think this couldn't be seem as a bug... but yeah, probably this is worth a compiler warning.



Sorry for the last reply, my answer didn't much make sense... I've read it quickly from my cellphone, didn't pay much attention.

> From: address@hidden
> To: address@hidden
> Date: Thu, 12 Apr 2012 15:44:13 +0100
> Subject: Re: [Tinycc-devel] Small curiosity, re operator precedence.
>
> I think the pre/post -- bit is ok though, not got a problem with that. What seems wrong is that TCC looks like it's giving the assignment priority over the comparison, and according to Wikipedia, the comparison sould have WAY more priority (9 on list, with assignment right down at 16 on list). It's not the location of the incrementor that matters, I think, but the assignment of the dereferenced value itself. It clearly works after the decrement as desired, and I like the way TCC works, it's just that it appears to be wrong, and I want clarification. As it is, I can't do this with GCC, which acts in accordance with what Wikipedia says about this...
>
> It seems to go beyond interpretations and implemetations of equal precedence, when one is as different from the other as 9 is from 16.
>
>
> Paulo Henrique Torrens <address@hidden> wrote:
> (12/04/2012 14:42)
>
> >This is a "feature". Ansi C says that the post and preincrements order are implementation defined. As it seems, there is something about the precedence order of the ==, that problably is some optimization. In order to achieve portability, separe it in some lines with the right order. But, yeah, that's no bug. :)
> >
> >
> >Lostgallifreyan <address@hidden> wrote:
> >(31/03/2012 09:47)
> >
> >>Please ignore any 'unsafe' aspect this example might have, it IS bounded in the context I'm using it in....
> >>
> >>while((*--o=*--i)!=92)
> >>
> >>I was tryign to reduce stuff, so came up with this to try:
> >>
> >>while(92!=*--o=*--i)
> >>
> >>TCC loves it, GCC hates it. :)
> >>
> >>GCC thinks it's doing this: while((92!=*--o)=*--i) and therefore an invalid lvalue, which makes sense given what Wikipedia has to say about precedence, but TCC (still using v0.9.24 here) seems to do what I was wondering if it might do, to compare the dereferenced value when it is copied. I reverted to the original form because the more compilers take my code the happier I am about it, but I'm interested. Is this a bug, or a feature? :)
> >>
> >
> >
> >Rudolph Loew, maker of the 48 bit LBA patch for W98, amongst other things, says this:
> >
> >It does appear to be a bug as far as I can tell.
> >They should not be reinterpreting the precedence even if the intent can be inferred.
> >
> >
> >I'm losing my limited power to code, I think my brain is dying. Not kidding. Not a coding problem so I'll say no more about it other than this: I cannot be trusted to fix the code, even if I try, especially as I don't know enough about TCC, and this needs to be looked at by someone working on TCC! You have a choice of two base assumptions: Rudolph Loew's brain is also dying, or there is a problem you need to check out.
> >
> >
> >_______________________________________________
> >Tinycc-devel mailing list
> >address@hidden
> >https://lists.nongnu.org/mailman/listinfo/tinycc-devel
> >
> >_______________________________________________
> >Tinycc-devel mailing list
> >address@hidden
> >https://lists.nongnu.org/mailman/listinfo/tinycc-devel
>
>
> _______________________________________________
> Tinycc-devel mailing list
> address@hidden
> https://lists.nongnu.org/mailman/listinfo/tinycc-devel