doc: fix header escaping regression (Issue #22065)

[rubys]

Remove backslashes in headers.

Fixes: #22065

Checklist

• make -j4 test (UNIX), or vcbuild test (Windows) passes
• commit message follows commit guidelines

[vsemozhetbyt]

Maybe just:

.replace(/\\(?=.)/g, ''),

[vsemozhetbyt]

Or, if we need to exclude escaped escapes:

.replace(/\\(?=[^\\])/g, ''),

[rubys]

jslint doesn't accept .replace(/\\(?=.)/g, ''),:

/Users/rubys/git/node/tools/doc/html.js
  202:29  error  Unescaped dot character in regular expression  node-core/no-unescaped-regexp-dot

I want to include escaped escapes.

Finally, lookahead isn't appropriate here. If we have an escaped escape, I want to emit a single backslash and then to not consider that backslash as an escape character. Consider four escapes in a row, the desired result would be two backslashes:

> "\\\\\\\\".replace(/\\(?=.)/g, '').length
1
> "\\\\\\\\".replace(/\\.{1}/g, (match) => match[1]).length
2

[vsemozhetbyt]

What about .replace(/\\(.{1})/g, '$1'),? Whould it be simpler and faster then function call and string index access?

[vsemozhetbyt]

BTW, we can disable linter rule for a line to simplify RegExp, but I am not sure what is more confusing:

// eslint-disable-next-line node-core/no-unescaped-regexp-dot
.replace(/\\./g, (match) => match[1]),

[vsemozhetbyt]

CI-lite: https://ci.nodejs.org/job/node-test-pull-request-lite-pipeline/404/

[vsemozhetbyt]

Not sure what is wrong with CI. Is it infrastructural fails?

[rubys]

+ sed '/^/\s*$/d'
sed: 1: "/^/\s*$/d": invalid command code \

Somebody either changed the version of sed or changed the build script. So, yes, I would say this looks like infrastructure fails.

Between the two, I would say that changing the build script seems considerably more likely. The second slash just looks outright wrong. Perhaps they meant to have a backslash escaping the next backslash so that they shell would pass sed a regular expression containing only a single backslash. If that is indeed the problem, it would seem amusing given the change that this particular pull request is addressing. :-)

[Trott]

Replay: https://ci.nodejs.org/job/node-test-pull-request-lite-pipeline/408/

[richardlau]

Looks to be a real failure:

17:58:34 /data/iojs/build/workspace/node-test-commit-linuxone/nodes/rhel72-s390x/tools/doc/html.js:203
17:58:34             position
17:58:34             ^^^^^^^^
17:58:34 
17:58:34 SyntaxError: Unexpected identifier
17:58:34     at new Script (vm.js:72:7)
17:58:34     at createScript (vm.js:234:10)
17:58:34     at Object.runInThisContext (vm.js:286:10)
17:58:34     at Module._compile (internal/modules/cjs/loader.js:657:28)
17:58:34     at Object.Module._extensions..js (internal/modules/cjs/loader.js:700:10)
17:58:34     at Module.load (internal/modules/cjs/loader.js:599:32)
17:58:34     at tryModuleLoad (internal/modules/cjs/loader.js:538:12)
17:58:34     at Function.Module._load (internal/modules/cjs/loader.js:530:3)
17:58:34     at Module.require (internal/modules/cjs/loader.js:637:17)
17:58:34     at require (internal/modules/cjs/helpers.js:20:18)
17:58:34 Building addon in /data/iojs/build/workspace/node-test-commit-linuxone/nodes/rhel72-s390x/test/addons/05_wrapping_c_objects
17:58:34 make[2]: *** [out/doc/api/path.html] Error 1

https://ci.nodejs.org/job/node-test-commit-linuxone/3483/nodes=rhel72-s390x/consoleFull

[richardlau]

Missing trailing comma?

[vsemozhetbyt]

CI-lite: https://ci.nodejs.org/job/node-test-pull-request-lite-pipeline/411/

[vsemozhetbyt]

Node.js Collaborators, please, add 👍 here if you approve fast-tracking.

[vsemozhetbyt]

It seems we have one more similar regression case: HTML entities verbatim rendering:

https://nodejs.org/download/nightly/v11.0.0-nightly20180803d68f946fe4/docs/api/inspector.html#inspector_event_lt_inspector_protocol_method_gt

Not sure if this should be fixed in this PR or in a new one.

[jasnell]

@rubys ... just a nit the prefix for the commit message here should be tools: rather than doc:

Also, just as a convention, it is helpful to include a [Squash] or [Fixup] prefix on commits that need to be squashed prior to landing.

[rubys]

My pull request for https://www.npmjs.com/package/mdast-util-to-hast has been accepted. Updating to the latest version of this dependency should fix everything. In fact, I should be able to back out some of the work that is currently committed. I'll run some tests and see what needs to be done from there.

[jasnell]

@rubys ... is this ready to go?

[rubys]

@jasnell: this PR is an accumulated set of workarounds to a remark bug, it certainly can go in now, but will all need to be backed out when these pull requests land: #22140 (comment)

[rubys]

CI: https://ci.nodejs.org/job/node-test-pull-request/16411/

[rubys]

My read is that the freebsd failure is unrelated. Unless there are any objections, I'll land this PR tomorrow.

[rubys]

Landed in 59f8276