Fix utime() syscall

[tiran]

Fix off by factor 1000: FS.utime() expects mtime and atime in seconds.

times == NULL is now correctly handled. utimes(path, NULL)
updates atime and mtime to current time.

https://nodejs.org/api/fs.html#fsutimespath-atime-mtime-callback

Fixes: #16458
Signed-off-by: Christian Heimes christian@python.org

[sbc100]

Thanks!

I'm curious to see if any tests fail.. if they don't it means we lack coverage.

[tiran]

Thanks!

I'm curious to see if any tests fail.. if they don't it means we lack coverage.

I compiled test_utime.c manually and patched the JS code. The test failed for me because errno was not reset to 0. I wonder if the test is actually executed...

The test file now resets errno, checks return values, and tests for utime(path, NULL)

[sbc100]

Hmm .. if this line is needed.. how was the test passing before?

[tiran]

I have no clue ...

[sbc100]

What test are you running that fails if you remove this line?

./tests/runner core2.test_utime works fine for on HEAD.

[tiran]

I executed the test manually with $ emcc -s NODERAWFS -o test_utime.js test_utime.c && node test_utime.js. The test failed with error message RuntimeError: Aborted(Assertion failed: !errno, at: test_utime.c,37,test). My guess is that some function earlier left errno != 0.

From man errno(3):

The value in errno is significant only when the return value of the call indicated an error (i.e., -1 from most system calls; -1 or NULL from most library functions); a function that succeeds is allowed to change errno. The value of errno is never set to zero by any system call or library function.

In other words, if you want to use errno for error checking, then you have to reset errno to 0 before a function call and check it immediately after the function call.

[sbc100]

Yes, I'm only confused about how that test was passing before.

It sounds like the tests passes because its not run with NODERAWFS today? i.e. ./tests/runner core2.test_utime does not set NODERAWFS. Perhaps we should annotate that test with also_with_noderawfs

[sbc100]

Seems a little odd to declare var mtime and var atime in both branches of the if/else. Maybe just declare them in one? (I'm not a JS expert but IIRC you just need one var decl for a give variable it it will be hoisted to function level).

[tiran]

I don't know any JS at all. :)

[syrusakbary]

mtime and atime should be declared on level up (var mtime, atime;).
Otherwise the FS.utime(path, atime, mtime) will not be able to access the variables on the scope.

[sbc100]

Is that true? I thought var declaration were hoisted to the function level in JS. This seems to work just fine:

function foo() {                                                                 
  if (0) {                                                                          
    var bar = 10;                                                                   
  } else {                                                                       
    bar = 7;                                                                     
  }                                                                                                
  console.error(bar);                                                            
} 

It looks odd but its saves a few bytes in code size.

[syrusakbary]

You are completely right! I got a bit confused with var and let :P

[sbc100]

lgtm % comments

[tiran]

Tests are failing.

The noderawfs layer defines the utime function as function() { fs.utimesSync.apply(void 0, arguments); }. According to https://nodejs.org/api/fs.html#fsutimessyncpath-atime-mtime and https://nodejs.org/api/fs.html#fsutimespath-atime-mtime-callback the utimesSync function accepts the time arguments as seconds, not milliseconds:

The atime and mtime arguments follow these rules:
Values can be either numbers representing Unix epoch time in seconds, Dates, or a numeric string like '123456789.0'.

Perhaps the values must be adjusted in the noderawfs layer instead?

[tiran]

Tests are no passing. I had to implement the same fix in wasmfs layer.

[sbc100]

lgtm! We can try to figure out why the missing errno = 0 was ok in the past but not any more as a followup.

[sbc100]

I'm curious how / why this was needed? Do we run test_utime under WASMFS? Or did you try running manually with this flag?

[tiran]

This was needed to implement utime(path, NULL) for wasmfs. The PR fixes utime(path, times) on noderawfs and implements utime(path, NULL) for all fs layers.

test_utime was failing on test_core.wasmfs, https://app.circleci.com/pipelines/github/emscripten-core/emscripten/19754/workflows/57871d6d-c5f8-4f0c-83ae-2566038f9724/jobs/513149

[sbc100]

At thanks.. i had forgotten that we can run the whole of the core tests suite in wasmfs mode