Kibana link checking is broken

[bmorelli25]

The Kibana link checker doesn't seem to be catching broken links.

docs/build_docs.pl

Lines 334 to 399 in 0cee4e4

	sub check_kibana_links {
	#===================================
	my $build_dir = shift;
	my $link_checker = shift;
	my $branch;
	say "Checking Kibana links";
	# ${baseUrl}guide/en/elasticsearch/reference/${urlVersion}/modules-scripting-expression.html
	# ${ELASTIC_WEBSITE_URL}guide/en/beats/filebeat/${DOC_LINK_VERSION}
	# ${ELASTIC_DOCS}search-aggregations-bucket-datehistogram-aggregation.html
	my $extractor = sub {
	my $contents = shift;
	return sub {
	while ( $contents =~ m!`(\$\{(?:baseUrl|ELASTIC_.+)\}[^`]+)`!g ) {
	my $path = $1;
	$path =~ s/\$\{(?:DOC_LINK_VERSION|urlVersion)\}/$branch/;
	$path
	=~ s!\$\{ELASTIC_DOCS\}!en/elasticsearch/reference/$branch/!
	|| $path =~ s!\$\{(?:baseUrl|ELASTIC_WEBSITE_URL)\}guide/!!;
	return ( split /#/, $path );
	}
	return;
	};
	};
	my $src_path = 'src/ui/public/documentation_links/documentation_links';
	my $legacy_path = 'src/legacy/ui/public/documentation_links/documentation_links';
	my $repo = ES::Repo->get_repo('kibana');
	my @branches = sort map { $_->basename }
	grep { $_->is_dir } $build_dir->subdir('en/kibana')->children;
	my $link_check_name = 'link-check-kibana';
	for (@branches) {
	$branch = $_;
	next if $branch eq 'current' || $branch =~ /^\d/ && $branch lt 5;
	say " Branch $branch";
	my $links_file;
	my $source = eval {
	$links_file = $src_path . ".js";
	$repo->show_file( $link_check_name, $branch, $links_file );
	} || eval {
	$links_file = $src_path . ".ts";
	$repo->show_file( $link_check_name, $branch, $links_file );
	} || eval {
	$links_file = $legacy_path . ".js";
	$repo->show_file( $link_check_name, $branch, $links_file );
	} || eval {
	$links_file = $legacy_path . ".ts";
	$repo->show_file( $link_check_name, $branch, $links_file );
	};
	die "failed to find kibana links file;\n$@" unless $source;
	$link_checker->check_source( $source, $extractor,
	"Kibana [$branch]: $links_file" );
	# Mark the file that we need for the link check done so we can use
	# --keep_hash with it during some other build.
	$repo->mark_done( $link_check_name, $branch, $links_file, 0 );
	}
	}

Here's an example of a broken link: elastic/kibana#63295.

[gtback]

I'm only starting to understand how Kibana link checking works. @nik9000 gave me an overview this morning. It might take me a little while to wrap my head around it all.

[gtback]

I figured out why this was broken (at least for elastic/kibana#63295):

The only links that are checked are those that are located in the root Doc Links file (which has moved and been renamed between versions of Kibana)

If you do something like this in a different file, and then build URLs in those files, those URLs won't be checked.

import { ELASTIC_WEBSITE_URL, DOC_LINK_VERSION } from 'ui/documentation_links';

This is what @lcawl is trying to fix in elastic/kibana#88107.

[gtback]

I'm tempted to close this, because I don't think looking through all of the Kibana code for links is feasible, but I'd like to track down why elastic/kibana#87721 wasn't failing as well.

[gtback]

I believe elastic/kibana#94274 (which prevents the doc_links_service.ts from being checked) is the ultimate root cause of this (along with elastic/kibana#88107: some of the links aren't in that file so aren't being checked).

Order of operations for resolving this:

• Fix Skip checking Kibana links that are on www.elastic.co but not docs links #2081.
• Ensure remove legacy doc links file kibana#94274 passes, then merge it in.
• Confirm Add Vega help link to DocLinksService kibana#87721 still fails. I merged in some of the changes from Add Vega help link to DocLinksService kibana#87721, so it's already failing, but merging in the latest should reduce the diff and it should still fail.
• Fix the link in Add Vega help link to DocLinksService kibana#87721 to confirm it passes.
• Close this issue

[gtback]

elastic/kibana#87721 is passing! 🙌🏻 So I'm confident we have this fixed. The actual link will get fixed once that's merged, but that's not a blocker here.