[Patch] PR java/22113: Fix buffer overflow in Java lexer for FP literals

[David Daney]

Ranjit Mathew wrote:
> Hi,
> 
>   This is a naive but effective patch for PR java/22113 where
> the 'literal_token' buffer of do_java_lex() in lex.c for reading
> in an FP literal was overflowing in cases like the Jacks testcase
> 3.10.2-round-6.
> 
> Tested on i686-pc-linux-gnu with Jacks included. Produces
> no new failures and fixes the recent FAIL-ing of 3.10.2-round-6
> on i686-pc-linux-gnu at least.
> 
> OK for mainline?
> 
> Thanks,
> Ranjit.
> 
> :ADDPATCH java:
> 
> 
> 
> ------------------------------------------------------------------------
> 
> Index: ChangeLog
> from  Ranjit Mathew  <rmathew@hotmail.com>
> 
> 	PR java/22113
> 	* lex.c (do_java_lex): Define MAX_TOKEN_LEN. Avoid overflowing
> 	`literal_token' for large numeric input tokens.
> 

What is the actual length of the longest of the tokens in question 
allowed by the JLS?

It seems if it is unbounded, that this is not the correct fix. 
Otherwise it seems unlikely that the upper bound is exactly 256.

Just my $0.02

David Daney