Bug 4329 – Do not show error messages that refer to __error

dmd should not output error messages like these: Error: function expected before (), not __error of type TOK149 I have never seen an error message that refers to __error, and that I wanted to see. They don't add any additional value, and are only (by definition?) caused by preceding, actual errors. They are only noise and don't add any additional value.

(In reply to comment #0) > dmd should not output error messages like these: > > Error: function expected before (), not __error of type TOK149 > > I have never seen an error message that refers to __error, and that I wanted to > see. They don't add any additional value, and are only (by definition?) caused > by preceding, actual errors. They are only noise and don't add any additional > value. __error is a new improvement to the compiler to supress spurious error messages. (in previous releases, it would have been 'int'). You should never see them. Please post any test cases which produce them.

Don, it seems to happen all the time. Normally I fix my compilation error right away, so I don't have a test case ready. See bug 4335 for one.

Copied from bug 4335 for convenience: --- import std.stdio; void main() { writeln("bug"); } --- t.d(3): Error: 'writeln' is not defined, perhaps you need to import std.stdio; ? t.d(3): Error: function expected before (), not __error of type _error_ (the first wrong error message is what bug 4335 is about)

BTW, note that this testcase is, unfortunately, D1 only.

It's possible that this is D1-only, I noticed the thing above only when working with D1 code.

Come on, fellas, you can do better than that. Here's 3 examples, and the last one produces a segfault. void bug4329() { with(great!().trepidation) return; try{ throw CautionToTheWind; } catch(LifeThreatening illness){} }

And here are another seven cases I found. TEST CASE B: -------------- void bug4329b() { create!().havoc!(); delete something; mixin(chaos); } bug.d(25): Error: template instance template 'create' is not defined bug.d(25): Error: undefined identifier create!().havoc bug.d(25): Error: __error isn't a template bug.d(26): Error: undefined identifier something bug.d(26): Error: cannot delete type _error_ bug.d(27): Error: undefined identifier chaos bug.d(27): Error: argument to mixin must be a string, not (__error) bug.d(27): Error: argument to mixin must be a string, not (__error) TEST CASE C: -------------- void bug4329c() { delete and!(crash); } bug.d(33): Error: identifier 'crash' is not defined bug.d(33): Error: and!(_error_) is not an lvalue bug.d(33): Error: cannot delete type void TEST CASE D: -------------- void bug4329d() { make!(mess); switch(2) { case 3:.. case 6.7: case closed: } } bug.d(37): Error: identifier 'mess' is not defined bug.d(37): Error: make!(_error_) has no effect bug.d(39): Error: cannot implicitly convert expression (6.7) of type double to i nt bug.d(39): Error: first case 3 is greater than last case __error bug.d(40): Error: undefined identifier closed bug.d(40): Error: case must be a string or an integral constant, not __error

This is a case I have found: void main() { int[] a = cast(int[])array(iota(10)); } dmd 2.047 prints: temp.d(2): Error: undefined identifier array temp.d(2): Error: undefined identifier iota Error: cannot cast _error_ to int[]

One case, dmd 2.047: void main() { auto x = foo.bar!(); } test.d(2): Error: undefined identifier foo test.d(2): Error: __error isn't a template

One case: void main() { enum Foo { E } Foo f; switch (f) { case 1: break; default: break; } } dmd 2.047 prints: test.d(5): Error: cannot implicitly convert expression (1) of type int to Foo test.d(5): Error: case must be a string or an integral constant, not __error

Another testcase: --- pragma (lib, l); void main() {} --- p.d(1): Error: undefined identifier l p.d(1): Error: pragma lib string expected for library name, not '__error'

struct Foo { int x; } void main() { Foo(5 ^^ 5); // line 5 } DMD 2.048 shows: test.d(5): Error: must import std.math to use ^^ operator test.d(5): Error: structliteral has no effect in expression (Foo((__error))) --------------------------- void foo() {} void main() { foreach (x; foo()[0 .. 1]) {} } DMD 2.048 shows: test.d(3): Error: foo() cannot be sliced with [] test.d(3): Error: foreach: _error_ is not an aggregate type

struct Vec(int N) { int[N] data; } void main() { Vec!(1)([1, 2]); } DMD 2.049 prints: test.d(5): Error: cannot implicitly convert expression ([1,2]) of type int[] to int test.d(5): Error: structliteral has no effect in expression (Vec((__error)))

*** Issue 2464 has been marked as a duplicate of this issue. ***

Original D1 case is patched here: https://github.com/donc/dmd/commit/1acd2b64b10c4a985a10750d5ad83a0b7e98332b Other cases patched for D2: https://github.com/donc/dmd/commit/3a512771e31de4fb8b4e806154b08766fd97aed1 and for D1: https://github.com/donc/dmd/commit/327afade5bfe0551176ee17ab6c991139dcb8058 Another test case from bug 5288: auto bug4329e(int z) { if (z) return undefined; else return 0; } fixed here: https://github.com/donc/dmd/commit/daff978816461602788f1cb396d9826918b6da21

https://github.com/D-Programming-Language/dmd/commit/3a512771e31de4fb8b4e806154b08766fd97aed1 https://github.com/D-Programming-Language/dmd/commit/daff978816461602788f1cb396d9826918b6da21