[NODERAWFS] Revert #18163

[kleisauke]

This reverts commit c614fa5.

---

Here's a standalone reproducer:

Before:

$ curl -LO https://github.com/llvm/llvm-test-suite/raw/main/SingleSource/Benchmarks/Misc-C++/Large/ray.cpp
$ em++ -O3 -sEXIT_RUNTIME -sNODERAWFS ray.cpp -o ray.js
$ node ray.js > x.ppm
$ sha256sum x.ppm
acaa1db08ee30dafda03b41c77466b1435fe8ffdd4d88e5a5876e87dff126f58  x.ppm
$ vips copy x.ppm x.png
VipsImage: memory area too small --- should be 262144 bytes, you passed 79485

After:

$ em++ -O3 -sEXIT_RUNTIME -sNODERAWFS ray.cpp -o ray.js
$ node ray.js > x.ppm
$ sha256sum x.ppm
cd6fa6fa75b4ba29077f7e4e531b96f17ff0b82f3c70df526c36f748d6a38c5c  x.ppm
$ vips copy x.ppm x.png

x.png

[sbc100]

I'm inclined to perhaps revisit #18163. It seems to me that if one uses NODERAWFS then overriding stdout, stderr, stdin, print, printErr should make just not effect the program stdout/stderr/stdin since the whole point of NODERAWFS is to directly map to node file I/O.

[kleisauke]

I'm inclined to perhaps revisit #18163. It seems to me that if one uses NODERAWFS then overriding stdout, stderr, stdin, print, printErr should make just not effect the program stdout/stderr/stdin since the whole point of NODERAWFS is to directly map to node file I/O.

I agree, let's revert #18163. In retrospect, it seems that library_tty.js is very tied to the browser, as it skips null terminators and always flushes with a newline.

The latest revision of this PR reverts #18163 in a trivial way.

[sbc100]

Can you update the PR title to refer to the change you are reverting?

[kleisauke]

Note that I set tty: false here to ensure that issue #22264 remains resolved.

$ emcc ioctl_test.c -lnodefs.js -lnoderawfs.js -sEXIT_RUNTIME
$ node ./a.out.js
pc-kaw
a.out.js: /dev/stdin not a tty: Not a tty

[kleisauke]

(I also set it for stdout and stderr, btw)

[sbc100]

Seems reasonable for now, but we should probably add a TODO there, and we should really be setting tty based on the underlying node isatty call for each descriptor.

[kleisauke]

Ah, I just noticed that isatty for stdio streams will return 0 and set errno to ENOTTY with this change, which is probably not what we want. I just pushed commit 88a5475 instead. Let's reopen #22264 after this change.

[kleisauke]

Can you update the PR title to refer to the change you are reverting?

Sure, updated to [NODERAWFS] Revert #18163.