Andrew Gierth <and...@tao11.riddles.org.uk> writes: > "Tom" == Tom Lane <t...@sss.pgh.pa.us> writes: > Tom> I also noticed that we'd been sloppy about making the file safe to > Tom> compile for both frontend and backend, so I cleaned that up.
> In a frontend, wouldn't it be more kosher to restore the previous SIGILL > handler rather than unconditionally reset it to SIG_DFL? If we had any other code that was setting the SIGILL trap, I might worry about that, but we don't. The whole thing is really a bit questionable to run in arbitrary environments -- for instance, it'd be pretty unsafe inside a threaded application. So if we had code in libpq or ecpg that computed CRCs, I'd be worrying about this approach quite a bit more. But it seems all right for current and foreseen uses. regards, tom lane