Re: egcs-1998-04-18: Typos in gcc/haifa-schedule.c, bad default specs

Re: egcs-1998-04-18: Typos in gcc/haifa-schedule.c, bad default specs

[Mike Stump]

> Date: Tue, 21 Apr 1998 16:29:30 +0200
> From: Franz Sirl <Franz.Sirl-kernel@lauterbach.com>

> >A patch was checked in for this Sunday.

> I think it would be a good idea if you answer such requests with a cvsweb
> link:

> This way people can easily be pointed to the correct patch and eventually
> taught to look first in cvsweb before posting bug reports ;-).

I think the status quo is fairly good.  The status quo is only a
simple, `it's fixed' message needs to be generated (easing the job of
the overworked people that make the compiler happen (Thanks Jeff)).

We should teach people to first cvs update, and retest before
reporting a bug.  This way, if there was such a fix in cvsweb, then
certainly the sources grabbed with cvs update would have the fix, and
certainly if they retested, the problem would not appear.  This is
easier on them, as then they don't have to read anything.  With your
proposal, they would.  If the problem isn't fixed, and not reported,
then reporting it is good.  If it is reported, then they can either
report it again (to let people know it is important), or just skip
sending in a report.

Requiring additional time from Jeff or Jim I think is a bad idea.  I
think their time is better spent doing what they do best.

Re: egcs-1998-04-18: Typos in gcc/haifa-schedule.c, bad default specs

[Jeffrey A Law]

  In message <199804211613.JAA20747@cygnus.com>you write:
  > Yeah, that would be true if all guys asking here would know about the
  > cvsweb possibility and how to use it.

  > The link on the web page is quite
  > hidden and in the INSTALL/FAQ it is still a "to be implemented" ite 
The best way to deal with these things is to help :-)

I agree the link on the web page isn't easily found (hell, it took
me a while to find it the first time though).

So why don't you send a patch to the web pages to improve the situation?

They can be checked out via "cvs co wwwdocs", you can send patches to
them just like any other code in the project.

I'm not an HTML or web page design guru.  In fact, I'm terrible at web
stuff, so suggestions/patches for the web page are greatly appreciated.

  > The
  > "look in the mail archive first before asking here" messages aren't very
  > helpful either, as long as the archive is not searchable via WebGlimpse (I
  > searched once an hour for a specific patch from H.J.Lu, back then the
  > archive covered only three months).
Yes, this is a problem.  Again, contributions for a search engine
on the mail archives would be greatly appreciated.  This may just be
a matter of tweaking the glimpse code.  I don't really know.

Note we have to be careful with indexing the mail archives -- the
whole recent archiver shutdown was due to indexing of the win32
archives sucking up every last byte of disk space on the web server :(


  > - mention cvsweb in INSTALL/* (I could prepare some text if you like)
INSTALL as it stands will probably disappear from the CVS repository,
it is redundant with the official web pages in wwwdocs.

We'll still include it with releases and maybe in snapshots, but
it needs to be revamped to work with the official web pages which
are kept elsewhere in the repository.

I just haven't figured out the best way to do this.

  > - bigger link to cvsweb with more comments on main web page
Awaiting patch :-)

  > - add an "X-Look-here-first-for-trivial-snapshot-patches: <cvsweb-URL>"
  > header to all mail sent out by majordomo ;-)
:-)  

  > - mention cvsweb in your replies from time to time
Good point.

  > - searchable mail archive
Desparately needed.

jeff

Re: egcs-1998-04-18: Typos in gcc/haifa-schedule.c, bad default specs

[Franz Sirl]

At 05:52 21.04.98 , Jim Wilson wrote:
>	The following have to be fixed to get gcc/haifa-sched.c to compile:
>
>A patch was checked in for this Sunday.

I think it would be a good idea if you answer such requests with a cvsweb
link:

< http://egcs.cygnus.com/cgi-bin/cvsweb/gcc/haifa-sched.c?r1=1.40&r2=1.41 >

or more general
< http://egcs.cygnus.com/cgi-bin/cvsweb/gcc/haifa-sched.c >

This way people can easily be pointed to the correct patch and eventually
taught to look first in cvsweb before posting bug reports ;-).

Ciao,
Franz.

Re: egcs-1998-04-18: Typos in gcc/haifa-schedule.c, bad default specs

[Franz Sirl]

At 17:17 21.04.98 , Jeffrey A Law wrote:
>
>
>  In message <199804211429.HAA14747@cygnus.com>you write:
>  > At 05:52 21.04.98 , Jim Wilson wrote:
>  > >	The following have to be fixed to get gcc/haifa-sched.c to compile:
>  > >
>  > >A patch was checked in for this Sunday.
>  > 
>  > I think it would be a good idea if you answer such requests with a cvsweb
>  > link:
>Generating these links means a rather significant amount of extra
>work to fix trivial problems.
>
>I think it should be sufficient to note that the problem is fixed;
>folks can/should browse cvsweb themselves to pick up the the fixes.

Yeah, that would be true if all guys asking here would know about the
cvsweb possibility and how to use it. The link on the web page is quite
hidden and in the INSTALL/FAQ it is still a "to be implemented" item. The
"look in the mail archive first before asking here" messages aren't very
helpful either, as long as the archive is not searchable via WebGlimpse (I
searched once an hour for a specific patch from H.J.Lu, back then the
archive covered only three months).

Remember, I don't want to put more load on you, I want to show you ways to
reduce the load.

What could/should be done:

- mention cvsweb in INSTALL/* (I could prepare some text if you like)
- bigger link to cvsweb with more comments on main web page
- add an "X-Look-here-first-for-trivial-snapshot-patches: <cvsweb-URL>"
header to all mail sent out by majordomo ;-)
- mention cvsweb in your replies from time to time
- searchable mail archive

Bye,
Franz.

Re: egcs-1998-04-18: Typos in gcc/haifa-schedule.c, bad default specs

[Jeffrey A Law]

  In message <199804211429.HAA14747@cygnus.com>you write:
  > At 05:52 21.04.98 , Jim Wilson wrote:
  > >	The following have to be fixed to get gcc/haifa-sched.c to compile:
  > >
  > >A patch was checked in for this Sunday.
  > 
  > I think it would be a good idea if you answer such requests with a cvsweb
  > link:
Generating these links means a rather significant amount of extra
work to fix trivial problems.

I think it should be sufficient to note that the problem is fixed;
folks can/should browse cvsweb themselves to pick up the the fixes.

jeff

Re: egcs-1998-04-18: Typos in gcc/haifa-schedule.c, bad default specs

[Jim Wilson]

	The following have to be fixed to get gcc/haifa-sched.c to compile:

A patch was checked in for this Sunday.

	GNU CPP version egcs-2.91.24 19980418 (gcc2 ss-980401 experimental) (i386
	Linux/ELF)
	cpp: Usage: ./cpp [switches] input output

And I just checked in a gcc.c patch for this problem.

Jim

egcs-1998-04-18: Typos in gcc/haifa-schedule.c, bad default specs

[Horst von Brand]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 1983 bytes --]

I'm using egcs-980321, binutils-2.9.0.3, glibc-2.0.92 (dated 980412) on i586

{,LIB}{C,CXX,F77}FLAGS = -O2 -fomit-frame-pointer -march=pentium
--enable-{shared,threads,haifa} i386-redhat-linux

The following have to be fixed to get gcc/haifa-sched.c to compile:

--- haifa-sched.c.dist	Sun Apr 19 13:12:00 1998
+++ haifa-sched.c	Sun Apr 19 13:53:11 1998
@@ -761,7 +761,7 @@
 static void find_pre_sched_live PROTO ((int));
 static void find_post_sched_live PROTO ((int));
 static void update_reg_usage PROTO ((void));
-static int queue_to_ready ((rtx [], int))
+static int queue_to_ready PROTO ((rtx [], int));
 
 void debug_ready_list PROTO ((rtx[], int));
 static void init_target_units PROTO (());

After fixing this, the resulting compiler crashes when trying to find out
if underscores are needed. Sure enough, the resulting cpp command line is
broken: 

./xgcc -v -B./ -S tmp-dum.c
Reading specs from ./specs
gcc version egcs-2.91.24 19980418 (gcc2 ss-980401 experimental)
 ./cpp -lang-c -v -isystem ./include -undef -D__GNUC__=2
 -D__GNUC_MINOR__=91 -D__ELF__ -Dunix -Dlinux -D__ELF__ -D__unix__
 -D__linux__ -D__unix -D__linux -Asystem(posix) -Di386 -Asystem(unix)
 -Acpu(i386) -Amachine(i386) __ -D__i386__ -Asystem(unix) -Acpu(i386)
 -Amachine(i386) __ tmp-dum.c /tmp/cca06809.i
GNU CPP version egcs-2.91.24 19980418 (gcc2 ss-980401 experimental) (i386
Linux/ELF)
cpp: Usage: ./cpp [switches] input output

I'm continuing the bootstrap after copying the specs file from 980321 over
(Yes, I know this is almost guaranteed to be wrong, but I just couldn't
find out where this stupid "__" comes from). BTW, the specs file made during
the building of a stage is equally broken, so I can't finish a normal
boostrap, and any checks will be suspect too :-(

For the time being, I'll consider egcs-1998-04-18 highly toxic.
-- 
Horst von Brand                             vonbrand@sleipnir.valparaiso.cl
Casilla 9G, Viña del Mar, Chile                               +56 32 672616