Calling Intl.v8BreakIterator makes the process crash

[targos]

node -p "new Intl.v8BreakIterator()"


#
# Fatal error in , line 0
# Failed to create ICU break iterator, are ICU data files missing?
#

==== C stack trace ===============================

 1: V8_Fatal
 2: v8::internal::BreakIterator::InitializeBreakIterator(v8::internal::Isolate*, v8::internal::Handle<v8::internal::String>, v8::internal::Handle<v8::internal::JSObject>, v8::internal::Handle<v8::internal::JSObject>)
 3: v8::internal::Runtime_CreateBreakIterator(int, v8::internal::Object**, v8::internal::Isolate*)
 4: 0x35e1cb4060bb
[1]    14295 illegal hardware instruction (core dumped)  node -p "new Intl.v8BreakIterator()"

This is the line where it happens:

node/deps/v8/src/i18n.cc

Line 914 in 64beab0

FATAL("Failed to create ICU break iterator, are ICU data files missing?");

I guess that the feature v8 is trying to load is not available with small-icu ? Anyway it seems wrong to experience a fatal error in this case.

[evanlucas]

Yea, I'm seeing this also. It works when built with full-icu, but fails with small-icu

[bnoordhuis]

/cc @nodejs/intl

[srl295]

V8 wanted this to be a fatal instead of just a soft error (exception or
something. )

Yes small Icu doesn't have the data needed for this undocumented feature.
Takes significant space- Use full Icu.

If this is an important feature lets get it added to ecma402

El martes, 29 de septiembre de 2015, Ben Noordhuis notifications@github.com
escribió:

/cc @nodejs/intl https://github.com/orgs/nodejs/teams/intl

—
Reply to this email directly or view it on GitHub
#3111 (comment).

[srl295]

Should we "delete" the Intl.v8BreakIterator property if we know at startup that it can't function? Thoughts anyone?

[caridy]

Does anyone know what Intl.v8BreakIterator does?

Maybe @ajklein knows more about this method.

[ajklein]

@nciric would know much better than I (this is the best I could find: https://code.google.com/p/v8-i18n/wiki/BreakIterator).

[nciric]

Break iterator does character/sentence segmentation in a locale friendly way. It relies on largish data files, and that's one of the reasons we added it to the Intl.* for Chrome/V8. Intl.v8BreakIterator is not part of ECMA 402 spec.

Feel free to remove support for it from node. V8 may needed it for some internal projects.

[Fishrock123]

Should we "delete" the Intl.v8BreakIterator property if we know at startup that it can't function? Thoughts anyone?

Nah just make it throw.

[srl295]

OK. Current thought:

• if the user sets a data path using env var or parameter (or in detect "full-icu" module #3460 full ICU is detected) , set process.icu_data_dir to that data path in node.cc
• in node.js:

       if (process.config.variables.icu_small && !process.icu_data_dir) {
          Intl.v8BreakIterator=function(){throw Error('Data not available…');}
      }

[caridy]

Does node exposes any other v8 specific functionality? I'm worry about portability of the code if you use v8BreakIterator().

[srl295]

@caridy no, nothing like that is exposed.