Re: include/sys/strace.h

[Christopher Faylor <cgf at redhat dot com>]

On Mon, Jan 28, 2002 at 09:33:46PM +0300, egor duda wrote:
>Hi!
>
>Monday, 28 January, 2002 Christopher Faylor cgf@redhat.com wrote:
>
>>>> >Chris, I'd actually kinda like to see this included, I can see it being
>>>> >handy from time to time.
>>>>
>>>> I don't agree.  It seems to me that this is easy enough to do with gdb.
>>>> I don't see any reason for it.
>>>
>>>The use case I see is when gdb hangs/crashes or the entire cygwin DLL
>>>hangs/crashes. In those instances, having a non-cygwin program that can
>>>monitor debug output would be highly useful.
>
>CF> However, your example doesn't make any sense to me.  I use the techniques
>CF> in how-to-debug-cygwin.txt and I've never had any problems.  There should
>CF> never issue of a hanging cygwin if you use these techniques.
>
>Ditto. Actually, strace is fully functional substitute for DbgView
>here. I know only of one more feature that may be useful: We may want
>to add one more type of *_printf, say private_printf, that is never
>used in main source tree. If someone has pinpointed the bug but
>suffers from "bug disappears under strace" problem, s/he just adds
>that private_printf()s to the code in question and sets strace mask to 
>catch only those "private" debug messages. That is, one can guarantee
>that strace won't be flooded with "ordinary" strace output. Does this
>make sense? I used this technique some time ago and can produce a
>patch if needed. 

I think this is what "minimal_printf" was supposed to accomplish but I
like the idea of a private_printf for this use.

We also need to resurrect the "attach to running process" capabilities
that you (Egor) submitted a patch for, IIRC.

I think I'll do that now.  I'm about to go into a boring meeting.

cgf