utils: fix partial writes with write_file_at_with_flags

[giuseppe]

@eriksjolund PTAL

[eriksjolund]

Just sharing an idea:

Maybe another way is to return

crun_make_error (err, EINVAL, "too large ...", name);

in case (len > INT_MAX) and then just return 0 for success.

The advantage would be that a caller could just check for 0 if the caller
does not care about the type of error.

Otherwise your version is fine too. I'm not sure what's best.

[giuseppe]

Just sharing an idea:

Maybe another way is to return

crun_make_error (err, EINVAL, "too large ...", name);

in case (len > INT_MAX) and then just return 0 for success.

The advantage would be that a caller could just check for 0 if the caller does not care about the type of error.

Otherwise your version is fine too. I'm not sure what's best.

realistically this is never going to happen, we only write very small files. But I'd prefer to not raise an error just because we picked the wrong type that cannot store a size_t. Alternatively, we could just not return this information and limit it to -1 for errors, 0 for success. And on success the whole buffer was written

[eriksjolund]

Alternatively, we could just not return this information and limit it to -1 for errors, 0 for success. And on success the whole buffer was written

Good idea. I like that as it makes it easier to read and understand the code.

[eriksjolund]

Alternatively, we could just not return this information and limit it to -1 for errors, 0 for success. And on success the whole buffer was written

Good idea. I like that as it makes it easier to read and understand the code.

I just want to add that it does not matter so much. Choose whatever is most convenient for you.

[giuseppe]

I'd prefer to not change the semantic too much in this PR, as we would need to check that every caller doesn't really use this information