Misguided message "Unable to locate an installed version of R"

[mcanouil]

Quarto displays "Unable to locate an installed version of R" when it can't find R which is fine.
Unfortunately, it does print the exact same message when R crashed and Quarto lost track of the binded R session.

See below a bad quality "screenshot" showing computations were running (custom knitr progression bar) until it R crashed.
(Quarto call is wrapped in a script, thus the other messages in the log file)

A more general message would be welcome that let know that "R might have crashed".

[cderv]

FWIW the message comes from

quarto-cli/src/core/knitr.ts

Lines 111 to 117 in 22c2d32

	export function rInstallationMessage(indent = "") {
	const lines = [
	"Unable to locate an installed version of R.",
	"Install R from " + colors.bold("https://cloud.r-project.org/"),
	];
	return lines.map((line: string) => `${indent}${line}`).join("\n");
	}

that we call in

quarto-cli/src/execute/rmd.ts

Lines 281 to 286 in 22c2d32

	async function printCallRDiagnostics() {
	const caps = await knitrCapabilities();
	if (!caps) {
	info("");
	info(rInstallationMessage());
	info("");

This diagnostic is called in several cases inside main callR function used to call in R process

quarto-cli/src/execute/rmd.ts

Lines 221 to 267 in 22c2d32

	try {
	const result = await execProcess(
	{
	cmd: [
	await rBinaryPath("Rscript"),
	resourcePath("rmd/rmd.R"),
	],
	cwd,
	stderr: quiet ? "piped" : "inherit",
	},
	input,
	"stdout>stderr",
	(output) => {
	if (outputFilter) {
	output = outputFilter(output);
	}
	return colors.red(output);
	},
	);
	if (result.success) {
	const results = await Deno.readTextFile(resultsFile);
	await Deno.remove(resultsFile);
	const resultsJson = JSON.parse(results);
	return resultsJson as T;
	} else {
	// quiet means don't print in normal cases, but
	// we still need to report errors
	if (quiet) {
	error(result.stderr || "");
	}
	if (reportError) {
	await printCallRDiagnostics();
	}
	return Promise.reject();
	}
	} catch (e) {
	if (reportError) {
	if (e?.message) {
	info("");
	error(e.message);
	}
	await printCallRDiagnostics();
	}
	return Promise.reject();
	}
	}

• Either when result.success is false, meaning when R executed but for some reason RScript returned an error code as result from it execProcess execution
• Either when execProcess for Rscript fails completely in DENO (we catch error after a try and print diagnostic)

Is there some context about how this happen ? It would help to have reproducible example

I tried one using

library(inline)
crashMe <- cfunction(body="::abort();")
crashMe()

which makes R crash

However, in quarto context and knitr context, this does not print the message you see

> quarto render index.qmd
Quarto version: 99.9.9
Error loading deno-dom-native: Could not open library: Le module spécifié est introuvable.

-- Searching for R binary --
Looking for 'Rscript' in QUARTO_R: undefined
Looking for 'Rscript' in R_HOME: C:/PROGRA~1/R/R-43~1.1
Looking for 'Rscript' in PATH.
Found in PATH at C:\Program Files\R\R-4.3.1\bin\x64\Rscript.exe

processing file: index.qmd
1/5
2/5 [unnamed-chunk-1]
3/5
4/5 [unnamed-chunk-2]
-- Searching for R binary --
Looking for 'Rscript' in QUARTO_R: undefined
Looking for 'Rscript' in R_HOME: C:/PROGRA~1/R/R-43~1.1
Looking for 'Rscript' in PATH.
Found in PATH at C:\Program Files\R\R-4.3.1\bin\x64\Rscript.exe

[NotebookContext]: Starting Cleanup
ERROR: Error
    at renderFiles (file:///C:/Users/chris/Documents/DEV_R/quarto-cli/src/command/render/render-files.ts:334:23)
    at eventLoopTick (ext:core/01_core.js:181:11)
    at async render (file:///C:/Users/chris/Documents/DEV_R/quarto-cli/src/command/render/render-shared.ts:98:18)
    at async Command.fn (file:///C:/Users/chris/Documents/DEV_R/quarto-cli/src/command/render/cmd.ts:212:26)
    at async Command.execute (file:///C:/Users/chris/Documents/DEV_R/quarto-cli/src/vendor/deno.land/x/cliffy@v0.25.4/command/command.ts:1790:7)
    at async quarto (file:///C:/Users/chris/Documents/DEV_R/quarto-cli/src/quarto.ts:135:3)
    at async file:///C:/Users/chris/Documents/DEV_R/quarto-cli/src/quarto.ts:167:5
NativeCommandExitException: Program "quarto.cmd" ended with non-zero exit code: 1.

This is full log with verbose mode https://quarto.org/docs/troubleshooting/#verbose-mode

Can you try share this too with bot stack trace and verbose mode ?

The message you see should be printed if knitrCapabilities() fails. In verbose mode, you can see we have two times the search for R binaries. I am curious of what this will print for you.

I wonder if this is not a problem of R crash, but more about R process not responding anymore , which make callR fails somehow in Quarto, and then knitrCapabilities() fails too somehow. I remember you told me about this, and mentioned very long computation.

[mcanouil]

When R runs out of memory and is killed by the system, there are no error code sent.
This is the "reason" in which it happened.

[cderv]

Can you try share this too with bot stack trace and verbose mode ?

OK, so this means no stack trace nor verbose mode throw when running quarto render in terminal ?

[mcanouil]

Probably not, I'll try when I have time to be sure.

[mcanouil]

Meanwhile, I make a reproducible example out of the following, here another time where Quarto message is not correct.

There is clearly an error, but still Quarto is wrongly saying "Unable to locate an installed version of R."

[cderv]

Thanks for keeping the example coming. Having some reproducible example will help us with different cases to check !

[cderv]

@mcanouil did you manage to make a reproducible example ?

We did improve the log message during debugging at #7013

Not exactly the same as the error message but already better.

[mcanouil]

I can reproduce 100% of the time on various environments but I can't share the data and was unable to make some alternative.

I suggest to close for now as completed via the improved messages.

[cderv]

I think I just need to understand in which part of the code an error like this #6929 (comment) can happen.

Is this an error while knitting ? Or an error while initializing R with renv for example (I see package discovery in #6929 (comment), so I assume renv)

[mcanouil]

The error occurs in the middle of the computation.
Schema:

1. first code cell works great
2. second code cell with memory heavy computation fails without error but crash R
3. third code cell warning/message by Quarto

I tried to replicate 2, but in every attempts R crash with an error message which aborts Quarto, so not the same thing.
It's possible that the crash occurs within an R package that has poor low-level code.