v2: Sitemaps are not always generated

[SamChou19815]

🐛 Bug Report

sitemap.xml is not always generated. It might be caused by #2443 and #2496 does not fix the issue for sitemap generation.

Here is the code that writes the sitemap file.

docusaurus/packages/docusaurus-plugin-sitemap/src/index.ts

Lines 38 to 43 in 6935b33

	const sitemapPath = path.join(outDir, 'sitemap.xml');
	fs.writeFile(sitemapPath, generatedSitemap, err => {
	if (err) {
	throw new Error(`Sitemap error: ${err}`);
	}
	});

Writing the file is async, and the process might already be killed before it was able to write the file. To confirm my hypothesis, I changed writeFile to writeFileSync and the problem seems to go away with my blog repo.

Have you read the Contributing Guidelines on issues?

Yes.

To Reproduce

The bug is not reproducible on every repo. I cannot reproduce it on the docusaurus v2 project, but I managed to reliably reproduce it for my blog repo (at least on my machine).

1. Clone https://github.com/SamChou19815/website
2. yarn
3. yarn workspace blog build
4. Observe packages/blog/build/sitemap.xml The file is empty.

Expected behavior

sitemap.xml is reliably generated.

Actual Behavior

sitemap.xml is not reliably generated.

Your Environment

• Docusaurus version used: v2-alpha.50
• Environment name and version (e.g. Chrome 78.0.3904.108, Node.js 10.17.0): Node 12
• Operating system and version (desktop or mobile): macOS 10.15.4

Reproducible Demo

(Paste the link to an example repo, including a siteConfig.js, and exact instructions to reproduce the issue.)

[lex111]

@yangshun so do we use synchronous write to sitemap file?

[yangshun]

Let's do that if we aren't already

…

On Sat, Apr 4, 2020, 3:08 PM Alexey Pyltsyn ***@***.***> wrote: @yangshun <https://github.com/yangshun> so do we use synchronous write to sitemap file? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#2516 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAKBCHILOAV5QJNUHNQQAPTRK3MFHANCNFSM4L3RLBLA> .

[lex111]

@SamChou19815 you don't want to do it? сб, 4 апр. 2020 г. в 11:42, Yangshun Tay <notifications@github.com>:

…

Let's do that if we aren't already On Sat, Apr 4, 2020, 3:08 PM Alexey Pyltsyn ***@***.***> wrote: > @yangshun <https://github.com/yangshun> so do we use synchronous write to > sitemap file? > > — > You are receiving this because you were mentioned. > Reply to this email directly, view it on GitHub > < #2516 (comment) >, > or unsubscribe > < https://github.com/notifications/unsubscribe-auth/AAKBCHILOAV5QJNUHNQQAPTRK3MFHANCNFSM4L3RLBLA > > . > — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#2516 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABBUIOZ3BEVCN56ZI4FARDDRK3XGLANCNFSM4L3RLBLA> .

[SamChou19815]

My concern is that the implementation detail of terminating the process has been leaked into the userland.

Let's imagine an example: suppose a user creates a plugin that generates a robots.txt file that includes all paths of docs but excludes everything else. If the user implements it in the same way we did currently (use writeFile with callback), then it will run into the same situation that the generation sometimes works and sometimes not.

I would be down to quickly switch writeFile to writeFileSync, but I think we need to think deeper about its broader implications.