I am porting some older gnu makefiles and linker scripts to a newer
version of GCC. I have one linker script that I don't see how to update.

SECTIONS
{
GROUP :
{
COMMON :
{
a.o
b.o
c.o
}
}
}

This generated a segfault until I removed the GROUP : { } encapsulation.
It links, but the order in which the symbols appear in the map is not
what I expect, and not the same as under the older tool.

simplest way to copy and rename all *.a files to *.b ???

Different people will have different answers. What I do is:

rename.pl 's/a$/b/' *.a

where rename.pl is the appended Perl script (it exists in many versions
around the net). This version is slightly modified by me from lwall's
original to print the successful renames.

karl

#!/usr/bin/env perl
# print changes --karl
#
# Revision 4.0 91/03/20 01:11:53 lwall
# 4.0 baseline.
#
# Revision 3.0.1.2 90/08/09 03:17:57 lwall
# patch19: added man page for relink and rename
'di';
'ig00';

simplest way to copy and rename all *.a files to *.b ???

Chris wrote:

I am not an awk syntax expert but AFAIK you don't want to split up the
awk program string into separate arguments. Quote the entire program
into one string argument. I think that is the root of your problem.
Your example shows an "if" in program argument one and the rest of the
program in program argument two.

I would do that this way using pattern-action pairs where the pattern
acts as an if statement. It just seems the most awk-ish way to me.

gawk '$8<0.001 {print "SNP:" $2 " P="$8}' Batch_Call_dev-bs.model.P1.model

I tested with this:

printf "1 one\n2 two\n3 three\n" | awk '$1<3 {print $0}'
1 one
2 two

But using the awk if control syntax in the awk program itself also
works okay. Here is a test using it:

When I run indent with various options I want against my source, it
does what I want but also messes with the placement of *s in pointer
types:

-int send_pkt(tpkt_t* pkt, void* opt_data);
-void dump(tpkt_t* bp);
+int send_pkt(tpkt_t * pkt, void *opt_data);
+void dump(tpkt * bp);

I know my placement of *s next to the type not the variable is
unconventional but how can I get indent to just leave them alone?
I've looked in the man page, the info page, and visited a half a dozen
pages that Google suggested and I can't find an option to do this.

I have source tree with many folders containing many .o files each. I
wish to partially link the contents of each folder so I have one
combined .o file (library?) for each folder. These combined (object
files?, libraries?) will go thru a final link step at a later stage in
the build.

Is there anything special about how this partial linking or can I just
do something like:

$(OBJS)=a.o b.o c.o
mylink.b : $(OBJS)
ld -o mylink.b $(OBJS)

Does building in this multi-tiered fashion result in a larger final
image, or can the final link still remove redundancies?

The goal is to satisfy both (a) and (b):

(a) the target(T) depends on a generated file(F) as well as
on the content of the file (file F contains a list of files).
(b) parallel make enabled

content = $(shell cat F)

T: F $(content)

F:
Rule to generate F

How do I do it?
TIA

James

I'm pulling my hairs trying to get 'sed' to work in a simple
makefile (under MingW/MSYS/Win32). I have the following in a
makefile:

VERSION = `grep WATTCP_VER_STRING inc/tcp.h | cut -d' ' -f4`
...
show_ver:
@echo This is version $(VERSION)

and '#define WATTCP_VER_STRING "2.2.11"' in inc/tcp.h.

This produces the expected ' This is version "2.2.11"'. But I really need
an output w/o the quotes. So I tried:
VERSION = `grep WATTCP_VER_STRING inc/tcp.h | cut -d' ' -f4 | sed -e /\\"//`

This does not work however. Anybody can explain me why? What should
the proper command be? I'm pretty lost here.

PS. The output from bash is:
/usr/bin/sh: -c: line 0: unexpected EOF while looking for matching ``'
/usr/bin/sh: -c: line 1: syntax error: unexpected end of file

--
Gisle V.

Hi!

I'm somewhat a newbie with diff, so maybe I'm missing that or there is a good
reason why it does not exist, but well, the gnu-website send me to this
mailaddress to ask for features, so I just do...

The current behaviour of diff (2.8.1) seems to be that if context between two
changes overlaps then only a single change hunk is created. Which is
certainly a good default. Still I found that there are situation where I
would prefer to create two change hunks instead which both use the same
contextlines. As the contextlines don't change this shouldn't conflict with
patch, right? So having an additional flag to control this would be great
(like --small_hunks or --no_shared_context)

Well, just a proposal, maybe someone likes it and codes it, or tells me about
a tool which already does that, or maybe someone tells me why this is a
really really stupid idea :-)

- Michael

* Ralf Wildenhues wrote on Fri, Aug 15, 2008 at 05:48:23AM CEST:

Erm, or just
grep --color=auto -e foo -e '$'

Seems to have been fixed since upstream.

Cheers,
Ralf