Hi, I'm pretty sure that's a separate bug, present with or without my patches. The code for TOK_CONTINUE (and TOK_BREAK, for that matter) needs to include a gen_vla_sp_restore, just as the code for TOK_GOTO does. (I think the code for TOK_FOR also needs to look like the code for a label to work with C9x initial declarations of VLAs, but that's a bit of a corner case...)
Anyway, your bug doesn't need the goto at all. I'll commit a fix in a minute. On Mon, Apr 27, 2015 at 2:58 PM, Sergey Korshunoff <[email protected]> wrote: > Additional vla_sp_save() is not a solution. Yes, it is, it just doesn't magically fix unrelated bugs :-) > if we add '{' after if(0) then there is no a "boom" even w/o > additional vla_sp_save(). Are you saying there's no segfault, or no "boom"? For me, the results are as follows: mob tcc, no { } block: segfaults mob tcc, { } block: segfaults 1..3 tcc, no { } block: segfaults 1..3 tcc, { } block: works 1..4 tcc, no { } block: works 1..4 tcc, { } block: works _______________________________________________ Tinycc-devel mailing list [email protected] https://lists.nongnu.org/mailman/listinfo/tinycc-devel
