Re: BUg in GNU make 3.75 (CYGWIN port)

Re: BUg in GNU make 3.75 (CYGWIN port)

[Earnie Boyd]

---N8TM@aol.com wrote:
8<
> I'm still mystified by these comments disparaging W9x.  My NT4/SP3
takes
> several times as long to run scripts as my W95 under cygwin, and W95
doesn't
> crash any more often than NT unless you insist on running without
USB or
8<

Well, I'll have to disagree with this.  NT for me has been about 5
times faster than Win95.  You must access some slow network links with
your scripts.  Although I have to say the NT3.51 was faster than
NT4SP3 which I've recently been upgraded to.

==
-                        \\||//
-------------------o0O0--Earnie--0O0o-------------------
--                earnie_boyd@yahoo.com               --
-- http://www.freeyellow.com/members5/gw32/index.html --
----------------------ooo0O--O0ooo----------------------

PS: Newbie's, you should visit my page.
_________________________________________________________
DO YOU YAHOO!?
Get your free @yahoo.com address at http://mail.yahoo.com

-
For help on using this list (especially unsubscribing), send a message to
"gnu-win32-request@cygnus.com" with one line of text: "help".

Re: BUg in GNU make 3.75 (CYGWIN port)

[Earnie Boyd]

---Glenn Spell <glenn@gs.fay.nc.us> wrote:
8<
> "I don't know whatall I hain't tried."
8<

How about one of the latest snapshots of the cygwin1.dll from
ftp://sourceware.cygnus.com/pub/cygwin/snapshots/
==
-                        \\||//
-------------------o0O0--Earnie--0O0o-------------------
--                earnie_boyd@yahoo.com               --
-- http://www.freeyellow.com/members5/gw32/index.html --
----------------------ooo0O--O0ooo----------------------

PS: Newbie's, you should visit my page.
_________________________________________________________
DO YOU YAHOO!?
Get your free @yahoo.com address at http://mail.yahoo.com

-
For help on using this list (especially unsubscribing), send a message to
"gnu-win32-request@cygnus.com" with one line of text: "help".

Re: BUg in GNU make 3.75 (CYGWIN port)

[Glenn Spell]

Earnie Boyd wrote:
> 
> How about one of the latest snapshots of the cygwin1.dll from
> ftp://sourceware.cygnus.com/pub/cygwin/snapshots/

Thanks. I didn't know about the new snapshots directory. I tried
the 0113 dll... no change in 'make' behavior. I also played around
with "forkchunk" and "chunksize"... I can't tell if that's
implemented yet or not. Since "chunksize" applies to the function
'fork_copy', I also tried adding a "Sleep (10)" to fork_copy...
that didn't help either.

-glenn

-- 
            Glenn Spell <glenn@gs.fay.nc.us>
         Fayetteville, North Carolina, U. S. A.
  ____________________________________________________
  ... blue skies ... happy trails ... sweet dreams ...
-
For help on using this list (especially unsubscribing), send a message to
"gnu-win32-request@cygnus.com" with one line of text: "help".

Re: BUg in GNU make 3.75 (CYGWIN port)

[N8TM]

In a message dated 1/12/99 4:28:46 AM Pacific Standard Time,
glenn@gs.fay.nc.us writes:

<< I've experienced problems with many different ports of 'make'.
  >>
Have you been using make-3.77?  It installs itself automatically if you have
one of those .exe-aware versions of install.exe.
-
For help on using this list (especially unsubscribing), send a message to
"gnu-win32-request@cygnus.com" with one line of text: "help".

Re: BUg in GNU make 3.75 (CYGWIN port)

[Glenn Spell]

N8TM@aol.com writes:
> glenn@gs.fay.nc.us writes:
> << I've experienced problems with many different ports of 'make'.
>   >>
> Have you been using make-3.77?  It installs itself automatically
> if you have one of those .exe-aware versions of install.exe.

Yes, I settled on a modified 3.77. Before Cygwin B20, I had modified
it to install ok. I also added some path-handling patches that I
extracted from the Cygnus modifications to 3.75.

While working on the Win95 B20 bug, I added a patch that Paul
Eggert posted to gnu.utils.bug back in July that changes the way
'make' handles the SIGCHLD signal. Instead of trying to keep its own
counter it now gets the status of its children from the OS. This
change will probably be included in make-3.78.

None of the above has helped the Win95 B20 bug in the least. More
recently, I've modified it to return a bogus load average of 3.0
and I set MAKEFLAGS to '-e -j 1 -l 0.0'. The '-l 0.0' coupled with
the bogus loadavg has "slowed it down" a little bit. This
has decreased the frequency of the crashes but it is still not
a suitable work-around. (BTW, the '-j 1' is default behavior,
added just for "good measure" and the '-e' was an afterthought
that cuts down on the work involved in enforcing my own
preferences with compiler flags.)

Actually, I modified it to run only one job while thinking that
it was running an infinite number of jobs and never checking
for the number of jobs or the load average. This 'make' actually
ran without ever crashing! But... it didn't keep track of the
processes and I would get errors about "no rule to make target
so-and-so" when said target was already being make by another
unfinished process.

"I don't know whatall I hain't tried."

-glenn

-- 
            Glenn Spell <glenn@gs.fay.nc.us>
         Fayetteville, North Carolina, U. S. A.
  ____________________________________________________
  ... blue skies ... happy trails ... sweet dreams ...
-
For help on using this list (especially unsubscribing), send a message to
"gnu-win32-request@cygnus.com" with one line of text: "help".

Re: BUg in GNU make 3.75 (CYGWIN port)

[N8TM]

In a message dated 1/13/99 2:42:48 PM Pacific Standard Time, dj@delorie.com
writes:

<< We all have 95 or 98 machines available for testing, but
 since NT is much more stable and reliable, we tend to use
 NT for the bulk of our development.  We don't get much
 work done when we have to reboot the machine all the time. >>

I'm still mystified by these comments disparaging W9x.  My NT4/SP3 takes
several times as long to run scripts as my W95 under cygwin, and W95 doesn't
crash any more often than NT unless you insist on running without USB or
running AOL and an egcs build at the same time, as I'm doing now.  make-3.77
is worth installing even on NT, but it's a necessity on W95.  I've had a few
egcs incremental patch problems on W95 which weren't replicated on NT, and the
egcs testsuite skips different sections on NT and W95, but I'll take the speed
of W95 ever NT any day.  Yes, linux is still faster than any Windoze, but I
find those version incompatibilities just as annoying as the Windoze ones.  It
does run egcs testsuite and struct OTOB.
-
For help on using this list (especially unsubscribing), send a message to
"gnu-win32-request@cygnus.com" with one line of text: "help".

Re: BUg in GNU make 3.75 (CYGWIN port)

[DJ Delorie]

> In a message dated 1/13/99 2:42:48 PM Pacific Standard Time, dj@delorie.com
> writes:
> 
> << We all have 95 or 98 machines available for testing, but
>  since NT is much more stable and reliable, we tend to use
>  NT for the bulk of our development.  We don't get much
>  work done when we have to reboot the machine all the time. >>
> 
> I'm still mystified by these comments disparaging W9x.

That was from personal experience.  *My* NT system is a better
development environment than the 95 side of the same machine.  My 95
tends to hang in places where NT won't, and debugging seems to work
better (for example, NT allows gdb to attach to running processes, but
I haven't been able to convince 95 to do that).

We also have some software debugging tools that only run under NT, due
to the way the underlying system is designed.  Task Manager is also a
helpful NT-only tool.

Your mileage may vary, of course.
-
For help on using this list (especially unsubscribing), send a message to
"gnu-win32-request@cygnus.com" with one line of text: "help".

BUg in GNU make 3.75 (CYGWIN port)

[Dominique PELLE]

On CYWIN_95-4.0 (Beta 20.1 release) the simple GNU Makefile
(GNU make 3.75) that follows does not work.  It works fine
with GNU make on Linux:

# --- cut here ---

all: /tmp/bar /tmp/foo ;

clean:
	rm -f /tmp/foo /tmp/bar


# This create problems on the CYWIN port of GNU make 3.75
/tmp/foo /tmp/bar:
	touch $@

# --- cut here ---


Here is how it fails:

$ make
make: *** No rule to make target `/tmp/bar', needed by `all'.  Stop.


The bug seems to occur when there are several targets
in the same rule AND when the target files are absolute pathnames

Everything is fine when relative pathnames are used instead
of absolute pathnames.  Everything is fine also when I replace
the rule:

/tmp/foo /tmp/bar:
	touch $@


... with 2 rules:

/tmp/foo:
	touch $@

/tmp/bar:
	touch $@



My host is:
BASH.EXE-2.02$ uname -a
CYGWIN_95-4.0 SPEEDY 20.1 (0.3/1/1) 1998-12-3 20:39:18 i586 unknown

My GNU make version is:
BASH.EXE-2.02$ make --version
GNU Make version 3.75, by Richard Stallman and Roland McGrath.
Copyright (C) 1988, 89, 90, 91, 92, 93, 94, 95, 96
        Free Software Foundation, Inc.
This is free software; see the source for copying conditions.
There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A
PARTICULAR PURPOSE.

Report bugs to <bug-gnu-utils@prep.ai.mit.edu>.
BASH.EXE-2.02$

-
For help on using this list (especially unsubscribing), send a message to
"gnu-win32-request@cygnus.com" with one line of text: "help".

Re: BUg in GNU make 3.75 (CYGWIN port)

[Glenn Spell]

Dominique PELLE writes>
> On CYWIN_95-4.0 (Beta 20.1 release) the simple GNU Makefile
> (GNU make 3.75) that follows does not work.  It works fine
> with GNU make on Linux:

Welcome to the Windows 95 Frustrated Users of Cygwin Klub.

> # This create problems on the CYWIN port of GNU make 3.75

I've experienced problems with many different ports of 'make'.

> The bug seems to occur when there are several targets
> in the same rule AND when the target files are absolute pathnames

I can do a 'make clean' on the Cygnus Cygwin top-level Makefile
and Windows 95 will crash everytime... every single time.
Windows will kill a child that is running the function
'spawn_guts' in 'spawn.cc'. And this is a forked child.
'make clean' is only 'make do-clean' and 'make local-clean'.
When I run 'make do-clean' and 'make local-clean' separately,
Windows *never* crashes... so it's not a long command line.

It does seem to always be 'sh' that is killed. It makes no
difference whether 'sh' is 'ash', 'bash', or whatever.

It's just a buggy Cygwin and none of the developers seem to
have any interest in working on Windows 95 problems. Even
the fine folks at Cygnus don't run Windows 95 (according to
one of their finest right here on this list).

-glenn

-- 
            Glenn Spell <glenn@gs.fay.nc.us>
         Fayetteville, North Carolina, U. S. A.
  ____________________________________________________
  ... blue skies ... happy trails ... sweet dreams ...
-
For help on using this list (especially unsubscribing), send a message to
"gnu-win32-request@cygnus.com" with one line of text: "help".

Re: BUg in GNU make 3.75 (CYGWIN port)

[DJ Delorie]

Glenn Spell wrote:
> Welcome to the Windows 95 Frustrated Users of Cygwin Klub.

Sigh, this is the first I've heard such disparaging remarks
about our support for 95.  I admit that cygwin under 95 doesn't
work as well as cygwin under NT, but a lot of that is due
to bugs in 95's support for certain win32 features that work
just fine in NT, and we have to put a lot of effort into
working around those bugs.

We do want 95 to be as usable as NT.  Perhaps you should
join the developer's list and help us identify and fix
the win95 problems?  We'll even help you get a local
cygwin build if you want to help us debug cygwin under 95.

> > # This create problems on the CYWIN port of GNU make 3.75
> 
> I've experienced problems with many different ports of 'make'.

Note that we're working on upgrading cygwin's make to 3.77, which
should fix many of the problems people have been reporting.

> It's just a buggy Cygwin and none of the developers seem to
> have any interest in working on Windows 95 problems.

Sorry, not true.

> Even
> the fine folks at Cygnus don't run Windows 95 (according to
> one of their finest right here on this list).

We all have 95 or 98 machines available for testing, but
since NT is much more stable and reliable, we tend to use
NT for the bulk of our development.  We don't get much
work done when we have to reboot the machine all the time.
I actually do most of my work on my Linux and SGI machines.
-
For help on using this list (especially unsubscribing), send a message to
"gnu-win32-request@cygnus.com" with one line of text: "help".

Re: BUg in GNU make 3.75 (CYGWIN port)

[David M. Smith]

In article <369B66FE.47E54769@delorie.com>, DJ Delorie <dj@delorie.com>
writes
>Note that we're working on upgrading cygwin's make to 3.77, which
>should fix many of the problems people have been reporting.

I have seen things like this several times in the list and it confuses
me.  I downloaded make-3.77.tar.gz unpacked it and built it.  The only
problem was the ".exe" one, which is easily sorted.  I am able to use
make-3.77 to rebuild Cygwin B20.1, it also builds RTEMS better than 3.75
ever did.

What work are you referring to?

Regards,
David
-- 
David M. Smith

Re: BUg in GNU make 3.75 (CYGWIN port)

[David M. Smith]

In article <369B66FE.47E54769@delorie.com>, DJ Delorie <dj@delorie.com>
writes
>Note that we're working on upgrading cygwin's make to 3.77, which
>should fix many of the problems people have been reporting.

I have seen things like this several times in the list and it confuses
me.  I downloaded make-3.77.tar.gz unpacked it and built it.  The only
problem was the ".exe" one, which is easily sorted.  I am able to use
make-3.77 to rebuild Cygwin B20.1, it also builds RTEMS better than 3.75
ever did.

What work are you referring to?

Regards,
David
-- 
David M. Smith