Incomplete switch statement causes espree to hang/crash

[xjamundx]

The following code will crash espree (so long as it's followed by a newline):

switch(x) {  case 'case':  break;

It's very easy to reproduce:

> require('espree').parse("switch(x) {  case 'case':  break;\n")
FATAL ERROR: CALL_AND_RETRY_LAST Allocation failed - process out of memory
[1]    63533 abort      node

[nzakas]

Yuck. I'm stuck away from a computer, any interest in investigating?

[xjamundx]

Yes I'll try to find out what's up sometime later today.

[xjamundx]

1. Verified it happens in esprima as well
2. When I run node --trace I get miles of

16:                *parseSourceElement+24(this=0x1a1581304121 <undefined>) {
  16:                } -> 0x1a1581304121 <undefined>
  16:                *parseSourceElement+24(this=0x1a1581304121 <undefined>) {
  16:                } -> 0x1a1581304121 <undefined>
  16:                *parseSourceElement+24(this=0x1a1581304121 <undefined>) {
  16:                } -> 0x1a1581304121 <undefined>
  16:                *parseSourceElement+24(this=0x1a1581304121 <undefined>) {
  16:                } -> 0x1a1581304121 <undefined>
  16:                *parseSourceElement+24(this=0x1a1581304121 <undefined>) {
  16:                } -> 0x1a1581304121 <undefined>
  16:                *parseSourceElement+24(this=0x1a1581304121 <undefined>) {
  16:                } -> 0x1a1581304121 <undefined>
  16:                *parseSourceElement+24(this=0x1a1581304121 <undefined>) {
  16:                } -> 0x1a1581304121 <undefined>
  16:                *parseSourceElement+24(this=0x1a1581304121 <undefined>) {
  16:                } -> 0x1a1581304121 <undefined>
  16:                *parseSourceElement+24(this=0x1a1581304121 <undefined>) {
  16:                } -> 0x1a1581304121 <undefined>
  16:                *parseSourceElement+24(this=0x1a1581304121 <undefined>) {
  16:                } -> 0x1a1581304121 <undefined>
  16:                *parseSourceElement+24(this=0x1a1581304121 <undefined>) {
  16:                } -> 0x1a1581304121 <undefined>
  16:                *parseSourceElement+24(this=0x1a1581304121 <undefined>) {
  16:                } -> 0x1a1581304121 <undefined>
  16:                *parseSourceElement+24(this=0x1a1581304121 <undefined>) {
  16:                } -> 0x1a1581304121 <undefined>
  16:                *parseSourceElement+24(this=0x1a1581304121 <undefined>) {
  16:                } -> 0x1a1581304121 <undefined>
  16:                *parseSourceElement+24(this=0x1a1581304121 <undefined>) {
  16:                } -> 0x1a1581304121 <undefined>
  16:                *parseSourceElement+24(this=0x1a1581304121 <undefined>) {
  16:                } -> 0x1a1581304121 <undefined>
  16:                *parseSourceElement+24(this=0x1a1581304121 <undefined>) {
  16:                } -> 0x1a1581304121 <undefined>
  16:                *parseSourceElement+24(this=0x1a1581304121 <undefined>) {
  16:                } -> 0x1a1581304121 <undefined>

[xjamundx]

Tracked down the bad code inside of parseSwitchCase() on line 4070:

    while (index < length) {
        if (match("}") || matchKeyword("default") || matchKeyword("case")) {
            break;
        }
        statement = parseSourceElement();
        consequent.push(statement);
    }

If you add console.log(consequent.length) you'll see that consequent grows forever.

[xjamundx]

Okay I think something like (if (!statement) { throwErrorTolerant({}, Message.UnclosedSwitch); } may work. Will test and get up a PR on monday. Also will check to see if Esprima has this fixed in their latest.

[nzakas]

Nice analysis! This might be a case where you need to see if lookahead is EOF and just abort.