Quoting Samuel Thibault (2014-02-22 16:53:11) > Justus Winter, le Sat 22 Feb 2014 16:17:13 +0100, a écrit : > > Quoting Samuel Thibault (2014-02-22 15:37:29) > > > Justus Winter, le Sat 22 Feb 2014 11:25:14 +0100, a écrit : > > > > * kern/ipc_mig.c (mig_strncpy): Do not zero-terminate the destination > > > > string. Return length of destination string. > > > > > > Well, that calls for some rationale :) > > > > > > I guess this is related with the other, MIG change? I.e. that this > > > modified version of kernel mig_strncpy depends on the fixed MIG > > > behavior? > > > > No. It's just that the version in mach did not return anything (and > > thus I believe that sending such a string from the kernel would never > > have worked). > > > The change in semantics is to make it behave like the version in the > > glibc (and like strncpy in general). > > Ok, but that contradicts the previous "OUGHT" comment :) And changing > the behavior means callers would get potentially broken. AIUI, the only > callers are the ones from mig-generated stubs ?
Yes. And furthermore, since mig_strncpy is supposed to return the string length (and it didn't), any use within the kernel was probably not working before. Justus
