Bug 190228 – [JSC] print() changes CRLF to CRCRLF on Windows

• Status: RESOLVED
• Resolution: FIXED
• Component: JavaScriptCore

[Ross Kirsling]

Repro:
1. Create test.js:
> print(function foo() {
> });

2. Run it (with WinCairo JSC) and view output with explicit carriage returns:
> jsc.exe test.js | cat -v

3. Notice that carriage returns inside the function are doubled:
> function fn() {^M^M
> }^M

---

Notes:
- This results in various test failures under the ChakraCore/test directory:
    Basics/with3.js.default
    Closures/closure-funcexpr-eval.js.default
    Function/CallerArgs.js.default
    Function/funcExpr.js.default
    Function/moreFuncExpr.js.default
    Function/stackargs.js.default
    Function/toString.js.default
    LetConst/l.js.default
    strict/05.arguments.js.default
- It appears that read/readFile also produce CRCRLF in function bodies; perhaps this is why there aren't even more failures (e.g. stress/jsc-read.js).

[Ross Kirsling]

Created attachment 351553 [details]
Patch

[Mark Lam]

Comment on attachment 351553 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=351553&action=review

> Source/JavaScriptCore/jsc.cpp:2236
> +    _setmode(1, _O_BINARY);
> +    _setmode(2, _O_BINARY);

Would it be possible to use _fileno(stdout) and _fileno(stderr) instead?  I think that those documents the intent better than 1 and 2.

[Ross Kirsling]

Created attachment 351555 [details]
Patch

[Mark Lam]

Comment on attachment 351555 [details]
Patch

r=me if the EWS bots are green.

[WebKit Commit Bot]

Comment on attachment 351555 [details]
Patch

Clearing flags on attachment: 351555

Committed r236827: <https://trac.webkit.org/changeset/236827>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.

[Radar WebKit Bug Importer]

<rdar://problem/44998734>