job control stuff in tracepoint.c

job control stuff in tracepoint.c

[Mark Kettenis]

Ok, I'm looking into cleaning up some host-dependent stuff;
autoconfigifying bits and so on.

I'm currently looking into stuff dealing with job-control signals
(SIGTSTP. sigprocmask, sigsetmask).  In that context we also do some
really wierd stuff with a macro called STOP_SIGNAL.  The reason for
its existence has been eliminated (broken job-control handling on the
Convex, support for which has been removed), yet it still exists.  I'd
like to eleminate STOP_SIGNAL (and revert to using SIGTSTP where
appropriate), but when I did that I discovered that STOP_SIGNAL is
used in tracepoint.c wheras it isn't defined at all, not even to
SIGTSTP.  Replacing #ifdef STOP_SIGNAL, with #ifdef SIGTSTP leads to
compilation failure since the variable job_control isn't declared.

I'm not quite sure what the code is trying to do.  Something with
redisplaying the prompt when running asynchronously?  Elena, can you
shed some light on this?

Mark

Re: job control stuff in tracepoint.c

[Elena Zannoni]

Mark Kettenis writes:
 > Ok, I'm looking into cleaning up some host-dependent stuff;
 > autoconfigifying bits and so on.
 > 
 > I'm currently looking into stuff dealing with job-control signals
 > (SIGTSTP. sigprocmask, sigsetmask).  In that context we also do some
 > really wierd stuff with a macro called STOP_SIGNAL.  The reason for
 > its existence has been eliminated (broken job-control handling on the
 > Convex, support for which has been removed), yet it still exists.  I'd
 > like to eleminate STOP_SIGNAL (and revert to using SIGTSTP where
 > appropriate), but when I did that I discovered that STOP_SIGNAL is
 > used in tracepoint.c wheras it isn't defined at all, not even to
 > SIGTSTP.  Replacing #ifdef STOP_SIGNAL, with #ifdef SIGTSTP leads to
 > compilation failure since the variable job_control isn't declared.
 > 
 > I'm not quite sure what the code is trying to do.  Something with
 > redisplaying the prompt when running asynchronously?  Elena, can you
 > shed some light on this?
 > 
 > Mark

Sure, I'll look into it. (Got a late start today).

Elena

Obsolete support in core; Was: job control stuff in tracepoint.c

[Andrew Cagney]

> The reason for
> its existence has been eliminated (broken job-control handling on the
> Convex, support for which has been removed), yet it still exists.


For the record, either way, it doesn't matter.  As far as I'm concerned, 
if someone finds a nasty hack that is there for an obsolete target, 
eliminating it is still an obvious fix.

There is a corollary, if someone finds nasty target/host dependant code 
then one way of fixing it is to obsolete that system :-)  Suggestions 
for host/target combinations welcome (ChrisF, sorry *-pe has already 
been nominated :-)

	Andrew

Re: job control stuff in tracepoint.c

[Elena Zannoni]

Mark Kettenis writes:
 > Ok, I'm looking into cleaning up some host-dependent stuff;
 > autoconfigifying bits and so on.
 > 
 > I'm currently looking into stuff dealing with job-control signals
 > (SIGTSTP. sigprocmask, sigsetmask).  In that context we also do some
 > really wierd stuff with a macro called STOP_SIGNAL.  The reason for
 > its existence has been eliminated (broken job-control handling on the
 > Convex, support for which has been removed), yet it still exists.  I'd
 > like to eleminate STOP_SIGNAL (and revert to using SIGTSTP where
 > appropriate), but when I did that I discovered that STOP_SIGNAL is
 > used in tracepoint.c wheras it isn't defined at all, not even to
 > SIGTSTP.  Replacing #ifdef STOP_SIGNAL, with #ifdef SIGTSTP leads to
 > compilation failure since the variable job_control isn't declared.
 > 
 > I'm not quite sure what the code is trying to do.  Something with
 > redisplaying the prompt when running asynchronously?  Elena, can you
 > shed some light on this?
 > 
 > Mark

OK. Nothing specifically required this code in async, except to keep
compatibility with the behavior in sync mode.  The code about SIGTSTP
in event-top.c is a clone (with async modifications) of what was in
top.c already.  

The fact that tracepoint.c doesn't know about job_control is because
it doesn't have an #include "terminal.h" (neither does inflow.c). The
reason I touched that code in tracepoint.c is because I needed to make
it work with the async model. Michael Snyder was the original author
of tracepoint.c.

(I looked at our internal repository history, and I didn't find
anyhting enlightening).

Actually, my guess is that this code in tracepoint.c was cloned from
top.c, command_line_input().
Given this, if you think it can be cleaned up, I would say go ahead.

Elena