[css-syntax] problem with parse component value

[tabatkins]

(migrated from mailing list)

Johannes Koch said:

I think there is a problem with parsing a component value (5.3.6) in the CR.

4. Reconsume the current input token. Consume a component value and let
   value be the return value. If nothing is returned, return a syntax error.

5. While the current input token is a <whitespace-token>, consume the
   next input token.

6. If the current input token is an <EOF-token>, return value.
   Otherwise, return a syntax error.

We need to consume the next input token before step 5. Otherwise at step
5, the current input token is the last token of the component value
consumed by 5.4.6. and cannot be either whitespace or EOF.

---

Tab Atkins said:

Ooh, very good catch. Fixed. (Rather than consuming the next token
eagerly, I just changed the condition to look at the next input
token.)

I'm also not sure why I wrote the first couple of steps so awkwardly.
I've instead change them as well to look at the next input token,
rather than eagerly consuming, looking at current, then reconsuming.

[tabatkins]

Not sure how to test this right now; the only consumer of this algorithm I can find is Web Manifest (to parse theme-color), and that's pretty hard to test.

(If anyone can help me find another API to test, I just need to verify that it works at all; the previous algorithm would always throw a syntax error.)

[tabatkins]

#3634, #3633, #3632, #3631, #3630, and #3629 all form a group of "your parsing algorithms were bad and you should feel bad". All of them are (a) difficult to test individually, since no APIs use them on their own yet (tho CSSOM should do so...), and (b) were broken to such an extent that, were any implementation to actually use them, it would immediately fail to parse basically all CSS.

As such, I'm declaring them all "Testing Unneeded".