miscellaneous thoughts from PavelVozenilek

[PavelVozenilek]

This fails:

pub fn main() -> %void {
  var x: u4 = undefined;
  x = @truncate(u4, 9999);
}

with strange error:

O:\zig0.1.1>zig build-exe b.zig
O:\zig0.1.1\b.zig:9:21: error: type '(integer literal)' has fewer bits than destination type 'u4'
  x = @truncate(u4, 9999);
                    ^

[PavelVozenilek]

@sizeof doesn't seem to work at all:

const io = @import("std").io;
pub fn main() -> %void {
  %%io.stdout.printf("{}\n", @sizeOf(u16));
}

gives error:

O:\zig0.1.1>zig build-exe b.zig
O:\zig0.1.1\b.zig:6:30: error: parameter of type '(integer literal)' requires comptime
  %%io.stdout.printf("{}\n", @sizeOf(u16));
                             ^

---

It is probably better to report 0.1.1 errors in one thread. Many previous issues have valuable discussions and flooding the tracker with every little problem would hide them.

[PavelVozenilek]

Automatic structure packing does not seem to work optimally:

const io = @import("std").io;

const xxx = struct {
 b1 : u2,
 x : u8,
 b2 : u2,
 y : u8,
 b3 : u2,
 b4 : u2,
 z : u32,
};

pub fn main() -> %void {
  var sz : u32 = @sizeOf(xxx);
  %%io.stdout.printf("{}\n", sz);
}

Invoking it gives 12 instead of expected 8. Other combinations give higher size too.

---

Edit: it also fails with no less-than-byte members:

const io = @import("std").io;

const xxx = struct {
  m1 : u16,
  m2 : u8,
  m3 : u32,
  m4 : u8,
};

pub fn main() -> %void {
  var sz : u32 = @sizeOf(xxx);
  %%io.stdout.printf("{}\n", sz);
}

Gives 12, should 8.

---

Btw, I think this feature (automatic rearangement of structure members) is mistake, wrong and confusing. In addition keyword packed is applied on opposite feature.

I'd once suggested solution here:
#488 (comment)

[PavelVozenilek]

Struct method cannot take non-const pointer. (Could be my misunderstanding.)

const xxx = struct {
  pub fn bar(self : &xxx) {}
};

pub fn main() -> %void {
  var x : xxx = undefined;
  x.bar();
}

fails with error:

O:\zig0.1.1>zig build-exe b.zig
O:\zig0.1.1\b.zig:9:3: error: expected type '&xxx', found 'xxx'
  x.bar();
  ^

When I add const to the pointer (&const xxx) it compiles. But I feel that it should be possible to pass self as mutable pointer.

[PavelVozenilek]

This compiles, in spite of struct method and struct member having the same name:

const xxx = struct {
  pub fn bar(self : &const xxx) {}
  bar : u8
};

pub fn main() -> %void {
}

But when I try to use the method it fails:

const xxx = struct {
  pub fn bar(self : &const xxx) {}
  bar : u8
};

pub fn main() -> %void {
  var x : xxx = undefined;
  x.bar();
}

with error:

O:\zig0.1.1>zig build-exe b.zig
O:\zig0.1.1\b.zig:10:4: error: type 'u8' not a function
  x.bar();
   ^

Something feels wrong.

[PavelVozenilek]

printf also doesn't like i128 type. This program:

const io = @import("std").io;


pub fn main() -> %void {
  var sz : i128 = undefined;

  sz = @minValue(u16);
  %%io.stdout.printf("{}\n", sz);
  sz = @minValue(i16);
  %%io.stdout.printf("{}\n", sz);
}

compiled but when I run the result executable it crashes.

[PavelVozenilek]

Zig requires initialization, even for empty structs:

const foo = struct {};

pub fn main() -> %void {

  var f : foo;
}

results in :

 
O:\zig0.1.1>zig build-exe b.zig
O:\zig0.1.1\b.zig:11:3: error: variables must be initialized
  var f : foo;
  ^

While empty struct does not make much of sense, asking for its initialization is IMO error.

[PavelVozenilek]

Docs states:

In Zig, structs, unions, and enums with payloads cannot be passed by value to a function.

However this compiles:

const foo = struct 
{
  x : i32,
  y : i32,
};

fn bar(f : foo) {}

pub fn main() -> %void {

  var f : foo = undefined;
  //bar(f);  // <<== it stops to compile when I uncomment this
}

This feels like a bug, the offending function has to be called for error to show.

[andrewrk]

Zig requires initialization, even for empty structs:

Can you make a separate issue for this one? I have a clear answer for you, and I want to be able to direct other people to the issue.

[andrewrk]

This feels like a bug, the offending function has to be called for error to show.

Same thing, can you make a separate issue? I have an answer and I want to be able to point back to the issue as a point of reference.

[PavelVozenilek]

Example copy pasted from the documentation works:

pub fn main() -> %void {
 const a  = true;
 const b = false;

 const result = if (a != b) 47 else 3089;
}

If I switch to variables it fails:

pub fn main() -> %void {
 var a  = true;
 var b = false;

 var result = if (a != b) 47 else 3089;
}

I get:

O:\zig0.1.1>zig build-exe b.zig
O:\zig0.1.1\b.zig:15:17: error: unable to infer expression type
 var result = if (a != b) 47 else 3089;
                ^

If I make only the result variable I get yet another strange error.

O:\zig0.1.1>zig build-exe b.zig
O:\zig0.1.1\b.zig:15:2: error: variable of type '(integer literal)' must be const or comptime
 var result = if (a != b) 47 else 3089;
 ^

This feels inconsistent and buggy.

[PavelVozenilek]

Initialization of struct member does not allow to use previously initialized member. I tried:

const io = @import("std").io;

const foo = struct 
{
  x : i32,
  y : i32,
};


pub fn main() -> %void {


  var bar : foo = foo {
    .x = 10,
    .y = .x + 10,
  };


  %%io.stdout.printf("x = {}, y = {}\n", bar.x, bar.y);
}

and got error:

O:\zig0.1.1>zig build-exe b.zig
O:\zig0.1.1\b.zig:15:10: error: invalid token: '.'
    .y = .x + 10,
         ^

Removing the leading dot didn't help.

[PavelVozenilek]

Possibly another problem with printf. I wanted to print pointer to a string argument.

This fails to compile:

const io = @import("std").io;

fn foo(comptime s: []const u8)
{
  var  ps : &const []const u8 = &s;
  %%io.stdout.printf("string = {}, ptr to string = {}\n", s, ps);
}

pub fn main() -> %void {
  foo("abc");
}

with error:

O:\zig0.1.1>zig build-exe b.zig
O:\zig0.1.1\lib\zig\std\fmt\index.zig:214:17: error: Unable to format type '&const []const u8'
                @compileError("Unable to format type '" ++ @typeName(T) ++ "'");
                ^
O:\zig0.1.1\lib\zig\std\fmt\index.zig:61:37: note: called from here
                    if (!formatValue(args[next_arg], context, output))
                                    ^
O:\zig0.1.1\lib\zig\std\io.zig:154:23: note: called from here
        _ = fmt.format(&context, printOutput, format, args);
                      ^
O:\zig0.1.1\lib\zig\std\io.zig:144:27: note: called from here
        %return self.print(format, args);
                          ^
O:\zig0.1.1\b.zig:9:21: note: called from here
  %%io.stdout.printf("string = {}, ptr to string = {}\n", s, ps);
                    ^
O:\zig0.1.1\b.zig:14:6: note: called from here
  foo("abc");
     ^
O:\zig0.1.1\lib\zig\std\fmt\index.zig:178:5: error: expected type 'bool', found 'void'
    switch (@typeId(T)) {
    ^
O:\zig0.1.1\lib\zig\std\fmt\index.zig:61:37: note: called from here
                    if (!formatValue(args[next_arg], context, output))
                                    ^
O:\zig0.1.1\lib\zig\std\io.zig:154:23: note: called from here
        _ = fmt.format(&context, printOutput, format, args);
                      ^
O:\zig0.1.1\lib\zig\std\io.zig:144:27: note: called from here
        %return self.print(format, args);
                          ^
O:\zig0.1.1\b.zig:9:21: note: called from here
  %%io.stdout.printf("string = {}, ptr to string = {}\n", s, ps);
                    ^
O:\zig0.1.1\b.zig:14:6: note: called from here
  foo("abc");

[PavelVozenilek]

@compileLog shortcoming: this code

const num = {
 var val : i32 = 99;
 @compileLog("comp time val = {}", val);
 val = val + 1;
 val
};

prints

| "comp time val = {}", 99
Instead of expected printf-like formatted result.

[PavelVozenilek]

Another problem with @compileLog:

This code:

const io = @import("std").io;

const num = {
 var val : i32 = 99;
 @compileLog("comp time val = {}", val);
 val = val + 1;
 val
};

pub fn main() -> %void {
  %%io.stdout.printf("In main runtime.\n");
}

does not invoke @compileLog at all. I sort of understand it - the num is not used and gets eliminated, together with its @compileLog. Only when I print the num value the @compileLog kicks in.

I think this is not what one expects. If some code contains active @compileLog it should be considered as "being used".

[PavelVozenilek]

Third problem with @compileLog: I have 3 such calls. The compiler prints the expected output and then emits 3 the same errors. Each error takes 3 lines. All together 12 lines.

Example:

O:\zig0.1.1>zig build-exe b.zig
| "comp time inside the main"
| "comp time val1 = {}", 99
| "comp time val2 = {}", 99
O:\zig0.1.1\b.zig:14:5: error: found compile log statement
    @compileLog("comp time inside the main");
    ^
O:\zig0.1.1\b.zig:6:2: error: found compile log statement
 @compileLog("comp time val1 = {}", val1);
 ^
O:\zig0.1.1\b.zig:22:2: error: found compile log statement
 @compileLog("comp time val2 = {}", val2);
 ^

O:\zig0.1.1>

With highlighting it is looks worse.

Why not something as:

| "comp time inside the main"     ( b.zig:14:5 )
| "comp time val1 = {}", 99          ( b.zig:6:2 )
| "comp time val2 = {}", 99        ( b.zig:22:2 )

[PavelVozenilek]

Inspired by "delete variable" thread I tried this:

const io = @import("std").io;

pub fn main() -> %void {

 var x : i32 = 1;
 x = undefined;
  %%io.stdout.printf("x = {}\n", x);
}

To my surprise it compiles, and prints out some garbage value.

Isn't this error, assignment of undefined at random place?

[PavelVozenilek]

Following fails to compile:

pub fn main() -> %void {
  var x = 1 << 1;
}

with error:

O:\zig0.1.1\b.zig:6:3: error: variable of type '(integer literal)' must be const or comptime
  var x = 1 << 1;
  ^

[PavelVozenilek]

Learning about @"foo" syntax. This compiles:

fn @"foo"() -> i32 { return 1; }

pub fn main() -> %void {

  %%io.stdout.printf("foo = {}\n", foo());
}

const io = @import("std").io;

It mixes @"foo" and foo and this seems wrong. I see no value in this, only potential confusion

[PavelVozenilek]

It seems that I can export C incompatible name using @"foo" syntax:

This compiles:

export fn @"fo o"() -> i32 { return 1; }

[PavelVozenilek]

I downloaded latest build ( 0.1.1.54a0db0d ) and copied the current Hello Word from docs:

const io = @import("std").io;

pub fn main() -> %void {
    // If this program is run without stdout attached, exit with an error.
    var stdout_file = %return io.getStdOut();
    const stdout = &stdout_file.out_stream;
    // If this program encounters pipe failure when printing to stdout, exit
    // with an error.
    %return stdout.print("Hello, world!\n");
}

I get this:

O:\zig0.1.1>zig build-exe b.zig
O:\zig0.1.1\b.zig:6:32: error: no member named 'out_stream' in struct 'File'
    const stdout = &stdout_file.out_stream;
                               ^

O:\zig0.1.1>

As an aside: Hello World is supposed to be the most simple example, something very gentle to newcomers.

[PavelVozenilek]

Crash in generated executable, version 0.1.1:

pub fn main() -> %void {
  comptime {
      @panic("comptime panic called");
  }
}

compiles OK. When result executable is invoked it prints

O:\zig0.1.1>b
comptime panic called
(stack trace unavailable for COFF object format)

and then crashes with standard Windows error dialog.

[PavelVozenilek]

Similarly it crashes when I invoke @panic in runtime.

pub fn main() -> %void {
  @panic("panic in runtime");
}

compiles and results in

O:\zig0.1.1>b
panic in runtime
(stack trace unavailable for COFF object format)

and the crash dialog.

[PavelVozenilek]

When I combine both compile time and run time @panic it doesn't compile:

pub fn main() -> %void {
comptime {
    @panic("comptime panic called");
}

  @panic("panic in runtime");
}

gives unexpected compiler error:

O:\zig0.1.1>zig build-exe b.zig
O:\zig0.1.1\b.zig:12:10: error: unreachable code
  @panic("panic in runtime");
         ^

O:\zig0.1.1>

I feel something is wrong here.

[PavelVozenilek]

So called "child type" of pointers does not work, neither in 0.1.1 nor in the head.

This line in docs:

assert((&u32).child == u32);
gives:

O:\zig-head>zig build-exe b.zig
O:\zig-head\b.zig:6:16: error: type '&u32' has no member called 'child'
  assert((&u32).child == u32);
               ^

I am also scratching my head what this feature means and where it could be used.

[andrewrk]

I welcome your feedback on Zig. In the future, I suggest this flow chart:

• Have you found a bug, a concrete proposal with code examples, or a use case that is not handled adequately?

  • File a new GitHub issue

• Do you have general feedback, or a discussion topic, or things you thought about while coding Zig?

  • Chat on IRC or send an email to the mailing list

• miscellaneous thoughts from PavelVozenilek #548 (comment) - I created ability to use @truncate on comptime integers #703 for you

• miscellaneous thoughts from PavelVozenilek #548 (comment) - duplicate of allow integer literals to be passed to var args parameters and parameters of type var #623

• miscellaneous thoughts from PavelVozenilek #548 (comment) - duplicate of re-order struct fields for better performance and memory usage #168

• miscellaneous thoughts from PavelVozenilek #548 (comment) - I created 0-byte structs do not allow taking non-const pointers as self parameter #704 for you

• miscellaneous thoughts from PavelVozenilek #548 (comment) - I created add a compile error note for field name shadowing struct member function #705 for you

• miscellaneous thoughts from PavelVozenilek #548 (comment) - this is fixed in master.

• miscellaneous thoughts from PavelVozenilek #548 (comment) - Yes, Zig requires initialization, even for empty structs. Feel free to open a new issue to discuss this.

• miscellaneous thoughts from PavelVozenilek #548 (comment) - Zig has lazy top level declarations. Unused things are not analyzed. Feel free to open a new issue to discuss this.

• miscellaneous thoughts from PavelVozenilek #548 (comment) - duplicate of unable to infer expression type when adding u32 to if-else number literals #137

• miscellaneous thoughts from PavelVozenilek #548 (comment) - Interesting proposal, would you like to file a separate issue?

• miscellaneous thoughts from PavelVozenilek #548 (comment) - This is a feature request for ability to print pointers. Please file a separate issue.

• miscellaneous thoughts from PavelVozenilek #548 (comment) - This is working as designed. Please file a separate issue.

• miscellaneous thoughts from PavelVozenilek #548 (comment) - This is working as designed. Please file a separate issue.

• miscellaneous thoughts from PavelVozenilek #548 (comment) - This is working as designed. Please file a separate issue.

• miscellaneous thoughts from PavelVozenilek #548 (comment) - This is working as designed. Please file a separate issue.

• miscellaneous thoughts from PavelVozenilek #548 (comment) - This is working as designed. Please file a separate issue.

• miscellaneous thoughts from PavelVozenilek #548 (comment) - duplicate of zig fmt: canonicalize symbol names #166

• miscellaneous thoughts from PavelVozenilek #548 (comment) - This is working as designed.

• miscellaneous thoughts from PavelVozenilek #548 (comment) - duplicate of Add documentation tests #465

• miscellaneous thoughts from PavelVozenilek #548 (comment) - I filed comptime panic does not emit a compile error #706 for you

• miscellaneous thoughts from PavelVozenilek #548 (comment) - working as designed

• miscellaneous thoughts from PavelVozenilek #548 (comment) - duplicate of comptime panic does not emit a compile error #706

• miscellaneous thoughts from PavelVozenilek #548 (comment) - duplicate of Add documentation tests #465