Fix memory leak when using continue or break statement with syntaxError

[gergocs]

This patch fixes #5062.

JerryScript-DCO-1.0-Signed-off-by: Gergo Csizi gergocs@inf.u-szeged.hu

[zherczeg]

Title: whit -> with

[zherczeg]

Does this patch increase memory consumption in the normal (no parse error) case?

[gergocs]

Yes it does. For example the regression-test-issue-1555.js test fails if the heap size is not increased.

[zherczeg]

Would it be possible to free the memory at its regular place as well?

[gergocs]

I don't think so. I have debugging this issue a very long time and didn't find any better methods to keep track of possible non freed pointers.

[zherczeg]

I mean when a pointer is not needed, it could be removed from the tracker and free it. I suspect we keep pointers too long around this way.

[zherczeg]

jerryscript/jerry-core/parser/js/js-parser.c

Line 3254 in 8d44eed

parser_raise_error (parser_context_t *context_p, /**< context */

The parser already does a lot of cleanup on throw. If we could put this branch list into some local context (a chain list with a head), we could cleanup it without complex structures.

[gergocs]

Currently it is a chain list not a complex structure and stored in the context to easily pass around.

[zherczeg]

Not the complexity, but it has while loops. The design is better if less scanning is needed.

Is it operating like a stack btw? So we usually remove items form the end?

[gergocs]

No, it's not like stack.