IRC, freenode, #hurd, 2013-01-18

<braunr> youpi: what would you think of adding a debug-related syscall to
  gnumach so that we have a printf-like tool even in situations where the
  code can't perform an rpc (i.e. glibc)
<youpi> could be useful indeed
<youpi> I've found myself having a hard time making some printfs from odd
  places of glibc :)
<braunr> i also suggest we make it unprivileged
<youpi> not enabled by default then
<youpi> otherwise it's an easy DoS
<braunr> well, for now, we don't care much about DoS, but we do care about
  debugging
<braunr> at least until the very core issues we have are understood and
  resolved
<youpi> I usually frown on debugging feature being too open
<braunr> me too
<youpi> you would always forget dropping one
<braunr> that's why i didn't suggest it earlier
<braunr> but i'm wasting too much time finding other decently effective
  ways

IRC, freenode, #hurd, 2013-01-19

<braunr> youpi: how about we build this debugging system call in debugging
  versions only ?
<braunr> i suppose you already use such versions for buildds anyway
<braunr> MACH_DEBUG is always enabled
<braunr> the debugging version only adds --enable-kdb if i'm right
<pinotree> check debian/rules
<braunr> that, and -fno-omit-frame-pointer

IRC, freenode, #hurd, 2013-01-21

<braunr> youpi: concerning gnumach, i've added a mach_print system call,
  with one argument (a null terminated string) to -dbg kernels
  (--enable-kbd)
<youpi> k
<braunr> if it's fine with you, i'll commit it too
<youpi> I'm fine

IRC, freenode, #hurd, 2013-04-07

<braunr> see http://www.sceen.net/~rbraun/mach_print/ for an example to use
  it

Makefile, mach print.S, main.c.

IRC, freenode, #hurd, 2013-07-01

<youpi> braunr: btw, we are missing the symbol in mach/Versions
<braunr> youpi: what symbol ?
<youpi> so the libc-provided RPC stub is not available
<youpi> mach_printf
<youpi> -f
<braunr> it's a system calll
<braunr> not exported
<youpi> s/RPC/system call/
<braunr> that's expected
<youpi> libc does provide stubs for system calls too
<braunr> yes but not for this one
<youpi> I don't see why we wouldn't want to include it
<youpi> ?! it does
<braunr> it's temporary
<braunr> oh
<braunr> there must be automatic parsing during build
<youpi> sure
<braunr> nice

<braunr> youpi: if we're going to make this system call exported by glibc,
  i should change its interface first
<braunr> it was meant as a very quick-and-dirty hack and directly accesses
  the caller's address space without going through a special copy-from-user
  function
<braunr> not very portable

IRC, freenode, #hurd, 2013-11-25

<teythoon> can we get the mach_printf function somewhere so that it's
  easier to use please ?
<braunr> well
<braunr> i'm not comfortable with that
<teythoon> y not ?
<braunr> i don't consider mach_print being part of the interface yet
<teythoon> ...
<braunr> it's really a debugging call
<braunr> not always available
<teythoon> so what, let it fail if it is not
<braunr> ok
<braunr> i should change it first
<braunr> pass a size argument
<braunr> and do a proper copyin
<teythoon> cool, thanks :)
<braunr> then we could add that printf function in libshouldbeinlibc i
  guess
<teythoon> that'd be nice, yes
<braunr> hm
<braunr> that would make only hurd servers able to use it though
<braunr> unless we preload it
<teythoon> well, i thought it's a staging area for libc ?
<braunr> in theory yes, in practice some functions have been stuck there
  for ages
<braunr> we'll discuss that with youpi and tschwinge 
<braunr> and pinotree