------- Comment #11 from tobi at gcc dot gnu dot org 2007-04-17 13:41 ------- (In reply to comment #10) > (In reply to comment #8) > > The patch looks good, though it would probably be a better idea to use > > tmpnam() > > to get the name for the temporary file. > > Why not. But I like the idea that it is predictable :)
Given that it's going to be deleted, it's not that much of a gain :) People always do stuff one wouldn't expect, so having filenames that could not possibly collide with user files is a good thing. > > A further thing one could do: instead of threatening "If you edit this, > > you'll > > get what you deserve.\n\n", one could actually verify that the MD5 sum > > matches > > the files contents. > > Too expensive! I'm against this additional cost (plus the developer cost of > writing and maintaining it). I agree that this is probably not worth it. > About testcases, you mean for the dejagnu testsuite? I have really no idea... Time to CC Janis? -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31587