1769755 - Improve decimal parsing code in TokenStream

Status: RESOLVED FIXED

(Core :: JavaScript Engine, task, P3)

Description

[Jan de Mooij [:jandem]]

I've been looking into using the double-conversion library for string-to-double conversion instead of dtoa.c and noticed some issues.

Comment 1

[Jan de Mooij [:jandem]]

Attachment: Bug 1769755 - Clean up TokenStream code for numbers starting with '0'. r?arai!

If the tokenizer sees a number like 0., it currently starts at the '.' and discards
the '0'. This relies on GetDecimalNonInteger => js_strtod_harder returning 0
for invalid strings (such as "." in this case). This patch fixes this so that we
handle 0.x similar to 1.x for example.

In GetDecimalNonInteger we can now assert the whole string was parsed.

This also fixes the decimalNumber comment:

• Case 3 no longer exists with this change, '0' is now handled similar to case 1.
• Case 2 didn't match the code after bug 1609547. Fix the comment.
• Case 4 (now case 3) didn't match the code after bug 1421400. Fix the comment.

Comment 2

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/12cafc8df1ca
Clean up TokenStream code for numbers starting with '0'. r=arai

Comment 3

[Atila Butkovits]

https://hg.mozilla.org/mozilla-central/rev/12cafc8df1ca