svn commit: r304928 - in head/lib/libc: amd64/sys i386/sys sys

[Konstantin Belousov]

On Mon, Aug 29, 2016 at 09:51:50AM -0700, John Baldwin wrote:
> On Monday, August 29, 2016 09:58:13 AM Konstantin Belousov wrote:
> > I dug into the ptrace(2) consumers, I found a lot of things using
> > it which I would not expect to use, besides usual suspects of gdb
> > lldb libunwind reptyr etc.  Most surprising was that even high-profile
> > consumers including gdb sometimes fail to check errno after PT_PEEK. On
> > the other hand, I did not found a case in gdb where errno is checked
> > after PT_PEEK but not zeroed before the syscall.
> 
> So the consumers are generally doing the right thing.
I only looked at gdb in details, other code is painful to read.

> Certainly I think having a C wrapper like this makes more sense than
> doing it all in assembly N times.  I would probably prefer to keep the
> manpage language the way it is though.
So do you want to revert man page bits ?

> 
> If we are really worried about compat, we could bump the ptrace symver
> and use the function above as the FBSD_1.0 version perhaps.
This does not solve anything.  Old-time compiled consumers would get
the expected interface, while the same consumers compiled anew get
incompatible interface.

Recent dirname/basename change demostrates the same problem.

On Mon, Aug 29, 2016 at 08:46:47PM +0300, Andrey Chernov wrote:
> Either we implement this wrapper or left all things as is, we need to
> document internal errno clearing additionally, to not make people wonder
> why errno becomes 0, probably with the mention that program should not
> relay on this obsoleted implementation feature.
It was done in r304928.  John want to revert this change, it seems.