www.digitalmars.com         C & C++   DMDScript  

digitalmars.D.learn - Compiler error on struct with invariant and postblitz

reply "Alex Khmara" <alex.khmara gmail.com> writes:
Please help: is it compiler bug or I'm doing something wrong?

This code:

module properties;

struct PropertyList {
<-->invariant() {
<--><-->assert(1);
<-->}

<-->this(this) {
<--><-->_props = [];
<-->}

<-->string[] _props;
}


gives compiler error:

Error: __result = this is not mutable
Error: __result = this is not an lvalue

Without invariant all works good.
Jan 01 2011
next sibling parent "Alex Khmara" <alex.khmara gmail.com> writes:
I forgot to mention - it's dmd 2.051
Jan 01 2011
prev sibling next sibling parent Jonathan M Davis <jmdavisProg gmx.com> writes:
On Saturday 01 January 2011 11:17:36 Alex Khmara wrote:
 Please help: is it compiler bug or I'm doing something wrong?
 
 This code:
 
 module properties;
 
 struct PropertyList {
 <-->invariant() {
 <--><-->assert(1);
 <-->}
 
 <-->this(this) {
 <--><-->_props = [];
 <-->}
 
 <-->string[] _props;
 }
 
 
 gives compiler error:
 
 Error: __result = this is not mutable
 Error: __result = this is not an lvalue
 
 Without invariant all works good.

It's hard to read what you're doing when have those extra arrows in there. In any case, create a bug report for it. There are several bugs at present with invariants not working because of things like a public member function being pure (that one you can fix by marking the invariant as pure) or returning a ref. My guess would be that the problem relates to bug http://d.puremagic.com/issues/show_bug.cgi?id=4867 which makes it so that postblit constructors cannot currently copy const or immutable objects, but I don't know, since problems with __result have to do with the return value of a function (which would be the result of postblit rather than what's passed in). So, just create a bug report on it and it can be sorted out with the other invariant, postblit, and const problems. For the moment, it probably means that you can't have both an invariant and a postblit constructor, which sucks, but until the bug is fixed, I don't see what else you can do. - Jonathan M Davis
Jan 01 2011
prev sibling parent "Alex Khmara" <alex.khmara gmail.com> writes:
On Sat, 01 Jan 2011 20:56:20 -0000, Jonathan M Davis <jmdavisProg gmx.com>  
wrote:

 On Saturday 01 January 2011 11:17:36 Alex Khmara wrote:
 Please help: is it compiler bug or I'm doing something wrong?

 This code:

 module properties;

 struct PropertyList {
 <-->invariant() {
 <--><-->assert(1);
 <-->}

 <-->this(this) {
 <--><-->_props = [];
 <-->}

 <-->string[] _props;
 }


 gives compiler error:

 Error: __result = this is not mutable
 Error: __result = this is not an lvalue

 Without invariant all works good.

It's hard to read what you're doing when have those extra arrows in there. In any case, create a bug report for it. There are several bugs at present with invariants not working because of things like a public member function being pure (that one you can fix by marking the invariant as pure) or returning a ref. My guess would be that the problem relates to bug http://d.puremagic.com/issues/show_bug.cgi?id=4867 which makes it so that postblit constructors cannot currently copy const or immutable objects, but I don't know, since problems with __result have to do with the return value of a function (which would be the result of postblit rather than what's passed in). So, just create a bug report on it and it can be sorted out with the other invariant, postblit, and const problems. For the moment, it probably means that you can't have both an invariant and a postblit constructor, which sucks, but until the bug is fixed, I don't see what else you can do. - Jonathan M Davis

Thank you. Sorry for arrows - Midnight Commander marks tabs with them, and I forgot to disable this. Filed bug: http://d.puremagic.com/issues/show_bug.cgi?id=5397
Jan 01 2011