This was a concern when we originally implemented the m[string(b)] optimization. If someone is simultaneously modifying b, then weird things happen during the lookup. But it won't crash anything, just maybe not return what you expect. And that's ok in cases of a data race, the spec makes no promises.
On Saturday, July 23, 2016 at 12:22:39 PM UTC-7, Axel Wagner wrote: > > The answer, I think, is yes. Because the only way, that any modification > of bytes can occur, is if there's a race and in the case of a race (hehe) > all bets are off anyway. I also see no reason for the scope of bytes to > have any influence on it, as races can also occur with variables in local > scope. If anything, escape analysis might. > > But to get a definite answer just try it out. Write the code, throw in > some concurrent modifications of bytes just for good measure (so the > compiler can't prove that no modification occurs) and look at the generated > code. > > On Sat, Jul 23, 2016 at 9:09 PM, Jan Mercl <0xj...@gmail.com <javascript:> > > wrote: > >> On Sat, Jul 23, 2016 at 9:03 PM T L <tapi...@gmail.com <javascript:>> >> wrote: >> >> > If "bytes" is a local variable, it is reasonable the compiler will do >> the optimization. >> > But how about if "bytes" is a global variable? >> >> Not sure what a global variable is. Go has universe scope, but user code >> cannot declare variables in that scope. Assuming a package scope is meant >> instead: such optimization might be possible Iff the compiler is able to >> prove that no one is ever mutating the slice. >> >> -- >> >> -j >> >> -- >> You received this message because you are subscribed to the Google Groups >> "golang-nuts" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to golang-nuts...@googlegroups.com <javascript:>. >> For more options, visit https://groups.google.com/d/optout. >> > > -- You received this message because you are subscribed to the Google Groups "golang-nuts" group. To unsubscribe from this group and stop receiving emails from it, send an email to golang-nuts+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
