[Qemu-devel] [PATCH] Remove -s flag in Makefile

[Qemu-devel] [PATCH] Remove -s flag in Makefile

[Glauber Costa]

install -s is pretty and cute, but makes the binary get into the build
directory without any debug information. This is bad, and makes glommer
sad. Also, tools that attempt to extract that data, like fedora debuginfo
extractor, will fail.

Signed-off-by: Glauber Costa <glommer@redhat.com>
---
 Makefile |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/Makefile b/Makefile
index 82fec80..859614b 100644
--- a/Makefile
+++ b/Makefile
@@ -251,7 +251,7 @@ endif
 install: all $(if $(BUILD_DOCS),install-doc)
 	mkdir -p "$(DESTDIR)$(bindir)"
 ifneq ($(TOOLS),)
-	$(INSTALL) -m 755 -s $(TOOLS) "$(DESTDIR)$(bindir)"
+	$(INSTALL) -m 755 $(TOOLS) "$(DESTDIR)$(bindir)"
 endif
 ifneq ($(BLOBS),)
 	mkdir -p "$(DESTDIR)$(datadir)"
-- 
1.6.0.6

Re: [Qemu-devel] [PATCH] Remove -s flag in Makefile

[Riku Voipio]

On Wed, Mar 25, 2009 at 03:45:14PM -0300, Glauber Costa wrote:
> install -s is pretty and cute, but makes the binary get into the build
> directory without any debug information. This is bad, and makes glommer
> sad. Also, tools that attempt to extract that data, like fedora debuginfo
> extractor, will fail.

Fully ACK, but you'll want to fix that in Makefile.target too.

> Signed-off-by: Glauber Costa <glommer@redhat.com>
> ---
>  Makefile |    2 +-
>  1 files changed, 1 insertions(+), 1 deletions(-)
> 
> diff --git a/Makefile b/Makefile
> index 82fec80..859614b 100644
> --- a/Makefile
> +++ b/Makefile
> @@ -251,7 +251,7 @@ endif
>  install: all $(if $(BUILD_DOCS),install-doc)
>  	mkdir -p "$(DESTDIR)$(bindir)"
>  ifneq ($(TOOLS),)
> -	$(INSTALL) -m 755 -s $(TOOLS) "$(DESTDIR)$(bindir)"
> +	$(INSTALL) -m 755 $(TOOLS) "$(DESTDIR)$(bindir)"
>  endif
>  ifneq ($(BLOBS),)
>  	mkdir -p "$(DESTDIR)$(datadir)"
> -- 
> 1.6.0.6
> 
> 

-- 
"rm -rf" only sounds scary if you don't have backups

Re: [Qemu-devel] [PATCH] Remove -s flag in Makefile

[Anthony Liguori]

Riku Voipio wrote:
> On Wed, Mar 25, 2009 at 03:45:14PM -0300, Glauber Costa wrote:
>   
>> install -s is pretty and cute, but makes the binary get into the build
>> directory without any debug information. This is bad, and makes glommer
>> sad. Also, tools that attempt to extract that data, like fedora debuginfo
>> extractor, will fail.
>>     
>
> Fully ACK, but you'll want to fix that in Makefile.target too.
>   

Does anyone object to this?  I know Debian also carries a similar patch.

Regards,

Anthony Liguori

>> Signed-off-by: Glauber Costa <glommer@redhat.com>
>> ---
>>  Makefile |    2 +-
>>  1 files changed, 1 insertions(+), 1 deletions(-)
>>
>> diff --git a/Makefile b/Makefile
>> index 82fec80..859614b 100644
>> --- a/Makefile
>> +++ b/Makefile
>> @@ -251,7 +251,7 @@ endif
>>  install: all $(if $(BUILD_DOCS),install-doc)
>>  	mkdir -p "$(DESTDIR)$(bindir)"
>>  ifneq ($(TOOLS),)
>> -	$(INSTALL) -m 755 -s $(TOOLS) "$(DESTDIR)$(bindir)"
>> +	$(INSTALL) -m 755 $(TOOLS) "$(DESTDIR)$(bindir)"
>>  endif
>>  ifneq ($(BLOBS),)
>>  	mkdir -p "$(DESTDIR)$(datadir)"
>> -- 
>> 1.6.0.6
>>
>>
>>     
>
>   

Re: [Qemu-devel] [PATCH] Remove -s flag in Makefile

[Lennart Sorensen]

On Wed, Mar 25, 2009 at 02:56:56PM -0500, Anthony Liguori wrote:
> Does anyone object to this?  I know Debian also carries a similar patch.

Makes sense since debian prefers to build binaries with debug info and
then strip them (unless DEB_BUILD_OPTIONS=nostrip is set) when generating
the .deb package.  Keeps things consistent.

-- 
Len Sorensen

Re: [Qemu-devel] [PATCH] Remove -s flag in Makefile

[Avi Kivity]

Anthony Liguori wrote:
> Riku Voipio wrote:
>> On Wed, Mar 25, 2009 at 03:45:14PM -0300, Glauber Costa wrote:
>>  
>>> install -s is pretty and cute, but makes the binary get into the build
>>> directory without any debug information. This is bad, and makes glommer
>>> sad. Also, tools that attempt to extract that data, like fedora 
>>> debuginfo
>>> extractor, will fail.
>>>     
>>
>> Fully ACK, but you'll want to fix that in Makefile.target too.
>>   
>
> Does anyone object to this?  I know Debian also carries a similar patch.

If people complain, we can add a ./configure flag for it.  But I doubt 
anyone will notice; the floppy days are gone.

-- 
I have a truly marvellous patch that fixes the bug which this
signature is too narrow to contain.

Re: [Qemu-devel] [PATCH] Remove -s flag in Makefile

[M. Warner Losh]

In message: <49CBCE91.1030307@redhat.com>
            Avi Kivity <avi@redhat.com> writes:
: Anthony Liguori wrote:
: > Riku Voipio wrote:
: >> On Wed, Mar 25, 2009 at 03:45:14PM -0300, Glauber Costa wrote:
: >>  
: >>> install -s is pretty and cute, but makes the binary get into the build
: >>> directory without any debug information. This is bad, and makes glommer
: >>> sad. Also, tools that attempt to extract that data, like fedora 
: >>> debuginfo
: >>> extractor, will fail.
: >>>     
: >>
: >> Fully ACK, but you'll want to fix that in Makefile.target too.
: >>   
: >
: > Does anyone object to this?  I know Debian also carries a similar patch.
: 
: If people complain, we can add a ./configure flag for it.  But I doubt 
: anyone will notice; the floppy days are gone.

But it is SOP to install stripped binaries....  I'm likely swimming
against the tide on this one...

Warner

Re: [Qemu-devel] [PATCH] Remove -s flag in Makefile

[Lennart Sorensen]

On Thu, Mar 26, 2009 at 01:02:58PM -0600, M. Warner Losh wrote:
> But it is SOP to install stripped binaries....  I'm likely swimming
> against the tide on this one...

It is also standard to use a packaging system. :)

-- 
Len Sorensen

Re: [Qemu-devel] [PATCH] Remove -s flag in Makefile

[Michael Jennings]

On Thursday, 26 March 2009, at 15:08:10 (-0400),
Lennart Sorensen wrote:

> > But it is SOP to install stripped binaries....  I'm likely swimming
> > against the tide on this one...
> 
> It is also standard to use a packaging system. :)

I advocate leaving as much choice/power/flexibility in the hands of
the end-user (or packager) as possible.  You can't "unstrip" a binary
installed with -s, but if you leave off the -s, the user can always
strip it later if they choose to do so.  If there were a configure
option, that would be ideal.  (It's a lot easier to add --enable-strip
to a spec file or control file than to have to maintain a patch from
release to release.)

Being that strip is lossy, it should not be hard-coded IMHO.  (Or, to
quote the sage of old, "That's one doodle that can't be undid,
home-skillet.")

Michael

-- 
Michael Jennings (a.k.a. KainX)  http://www.kainx.org/  <mej@kainx.org>
Linux Server/Cluster Admin, LBL.gov       Author, Eterm (www.eterm.org)
-----------------------------------------------------------------------
 "Persistence trumps talent and looks every time."      -- Aaron Brown

Re: [Qemu-devel] [PATCH] Remove -s flag in Makefile

[M. Warner Losh]

In message: <20090326190810.GH3795@csclub.uwaterloo.ca>
            lsorense@csclub.uwaterloo.ca (Lennart Sorensen) writes:
: On Thu, Mar 26, 2009 at 01:02:58PM -0600, M. Warner Losh wrote:
: > But it is SOP to install stripped binaries....  I'm likely swimming
: > against the tide on this one...
: 
: It is also standard to use a packaging system. :)

It is also standard to strip before the packaging system gets
involved...

Warner

Re: [Qemu-devel] [PATCH] Remove -s flag in Makefile

[Lennart Sorensen]

On Thu, Mar 26, 2009 at 01:30:39PM -0600, M. Warner Losh wrote:
> It is also standard to strip before the packaging system gets
> involved...

Well on debian (and the many derivatives) it is the package building
step that does the stripping if it isn't told not to, which makes it
easy to build a package with debug symbols if you need one, without
having to know how the actual make files for the code work.  Having to
edit a makefile to enable debuging is annoying.

-- 
Len Sorensen

Re: [Qemu-devel] [PATCH] Remove -s flag in Makefile

[M. Warner Losh]

In message: <20090326194839.GI3795@csclub.uwaterloo.ca>
            lsorense@csclub.uwaterloo.ca (Lennart Sorensen) writes:
: On Thu, Mar 26, 2009 at 01:30:39PM -0600, M. Warner Losh wrote:
: > It is also standard to strip before the packaging system gets
: > involved...
: 
: Well on debian (and the many derivatives) it is the package building
: step that does the stripping if it isn't told not to, which makes it
: easy to build a package with debug symbols if you need one, without
: having to know how the actual make files for the code work.  Having to
: edit a makefile to enable debuging is annoying.

The whole world isn't Linux. :)

Warner

Re: [Qemu-devel] [PATCH] Remove -s flag in Makefile

[Lennart Sorensen]

On Thu, Mar 26, 2009 at 02:07:32PM -0600, M. Warner Losh wrote:
> The whole world isn't Linux. :)

Nooo?  Really?  Darn.  Must be a defect of my world then. :)

Well as someone else said, calling configure with an argument can make
install still call with '-s'.  I suppose one could even make a make
target that adds the -s option while installing.  It is easier to add
the -s option than to remove it when it is hardcoded in after all.

-- 
Len Sorensen

Re: [Qemu-devel] [PATCH] Remove -s flag in Makefile

[Riku Voipio]

On Thu, Mar 26, 2009 at 01:30:39PM -0600, M. Warner Losh wrote:
> In message: <20090326190810.GH3795@csclub.uwaterloo.ca>
>             lsorense@csclub.uwaterloo.ca (Lennart Sorensen) writes:
> : On Thu, Mar 26, 2009 at 01:02:58PM -0600, M. Warner Losh wrote:
> : > But it is SOP to install stripped binaries....  I'm likely swimming
> : > against the tide on this one...
> : 
> : It is also standard to use a packaging system. :)

> It is also standard to strip before the packaging system gets
> involved...

Stnadard where? In automake world it's not. There is a
install-strip target available on automake.

Picking up a random non-automaked source (rsync), and behold,
no strip in install stage. To check if this puts the non-linux
people in deep misery, nope. The FreeBSD port copes fine
by adding a single strip command in postinstall: target.

Unconditional stripping is bad, mmkay?

-- 
"rm -rf" only sounds scary if you don't have backups

Re: [Qemu-devel] [PATCH] Remove -s flag in Makefile

[M. Warner Losh]

In message: <20090326210909.GA19263@kos.to>
            Riku Voipio <riku.voipio@iki.fi> writes:
: On Thu, Mar 26, 2009 at 01:30:39PM -0600, M. Warner Losh wrote:
: > In message: <20090326190810.GH3795@csclub.uwaterloo.ca>
: >             lsorense@csclub.uwaterloo.ca (Lennart Sorensen) writes:
: > : On Thu, Mar 26, 2009 at 01:02:58PM -0600, M. Warner Losh wrote:
: > : > But it is SOP to install stripped binaries....  I'm likely swimming
: > : > against the tide on this one...
: > : 
: > : It is also standard to use a packaging system. :)
: 
: > It is also standard to strip before the packaging system gets
: > involved...
: 
: Stnadard where? In automake world it's not. There is a
: install-strip target available on automake.

All the world isn't automake.

I guess I've been exposed to install -s by default systems since I
started using Unix on both the 4.2BSD VAX we had as well as the Sys
III (later SYS V) AT&T hardware.  And through the years on SunOS,
Solaris, HPUX, AIX, Linux, etc.

I guess this boils down to "It's how its been done since the early
1980's at least" but I do know times change.

: Picking up a random non-automaked source (rsync), and behold,
: no strip in install stage. To check if this puts the non-linux
: people in deep misery, nope. The FreeBSD port copes fine
: by adding a single strip command in postinstall: target.
: 
: Unconditional stripping is bad, mmkay?

Of course.  That's why most systems have a flag called STRIP that's
passed to install so that users can turn it on or off, but it defaults
to on.

Warner

Re: [Qemu-devel] [PATCH] Remove -s flag in Makefile

[Riku Voipio]

On Thu, Mar 26, 2009 at 03:16:38PM -0600, M. Warner Losh wrote:
> I guess this boils down to "It's how its been done since the early
> 1980's at least" but I do know times change.

Maybe you have heard about the new kids called GNU?

http://www.gnu.org/prep/standards/standards.html#Standard-Targets

I don't think qemu is following GNU guidelines, but these are
pretty much the makefile targets *most* users expect these days.

> : Unconditional stripping is bad, mmkay?

> Of course.  That's why most systems have a flag called STRIP that's
> passed to install so that users can turn it on or off, but it defaults
> to on.

Personally, this solution would work for me. Not that my random
sampling from the debian archive found any application that uses
STRIP in that way - when STRIP was defined, it is usually "strip".
Including qemu btw...

-- 
"rm -rf" only sounds scary if you don't have backups

Re: [Qemu-devel] [PATCH] Remove -s flag in Makefile

[M. Warner Losh]

In message: <20090326221122.GA25747@kos.to>
            Riku Voipio <riku.voipio@iki.fi> writes:
: On Thu, Mar 26, 2009 at 03:16:38PM -0600, M. Warner Losh wrote:
: > I guess this boils down to "It's how its been done since the early
: > 1980's at least" but I do know times change.
: 
: Maybe you have heard about the new kids called GNU?

Nah.  They are just a fad that will fade away soon enough :)

: http://www.gnu.org/prep/standards/standards.html#Standard-Targets
: 
: I don't think qemu is following GNU guidelines, but these are
: pretty much the makefile targets *most* users expect these days.

These are just the coding standards for FSF projects.  But that does
have a lot of influence on other users in the worlds...

: > : Unconditional stripping is bad, mmkay?
: 
: > Of course.  That's why most systems have a flag called STRIP that's
: > passed to install so that users can turn it on or off, but it defaults
: > to on.
: 
: Personally, this solution would work for me. Not that my random
: sampling from the debian archive found any application that uses
: STRIP in that way - when STRIP was defined, it is usually "strip".
: Including qemu btw...

Yea, don't much work with debians.

Warner

Re: [Qemu-devel] [PATCH] Remove -s flag in Makefile

[Gerd Hoffmann]

M. Warner Losh wrote:
> In message: <20090326190810.GH3795@csclub.uwaterloo.ca>
>             lsorense@csclub.uwaterloo.ca (Lennart Sorensen) writes:
> : On Thu, Mar 26, 2009 at 01:02:58PM -0600, M. Warner Losh wrote:
> : > But it is SOP to install stripped binaries....  I'm likely swimming
> : > against the tide on this one...
> : 
> : It is also standard to use a packaging system. :)
> 
> It is also standard to strip before the packaging system gets
> involved...

No.  binutils and gdb can handle debuginfo being stored in separate
files these days.  The usual packaging procedure is to install
unstripped, then split away debuginfo to separate files for a separate
package, so you end up with one $software and one $software-debuginfo
package.  By default only $software is installed, but if needed you can
install get debuginfo too without having to compile the package manually.

I'd tend to keep the default behavior to install stripped.  There should
be an easy way out for packaging tools, they should be able to override
the default without having to patch the Makefile(s).  Some configure
script option should do the trick.

cheers,
  Gerd

Re: [Qemu-devel] [PATCH] Remove -s flag in Makefile

[Robert Riebisch]

Gerd Hoffmann wrote:

> No.  binutils and gdb can handle debuginfo being stored in separate
> files these days.  The usual packaging procedure is to install
> unstripped, then split away debuginfo to separate files for a separate
> package, so you end up with one $software and one $software-debuginfo
> package.  By default only $software is installed, but if needed you can
> install get debuginfo too without having to compile the package manually.
> 
> I'd tend to keep the default behavior to install stripped.  There should
> be an easy way out for packaging tools, they should be able to override
> the default without having to patch the Makefile(s).  Some configure
> script option should do the trick.

Agreed.

Robert Riebisch
-- 
BTTR Software
http://www.bttr-software.de/

Re: [Qemu-devel] [PATCH] Remove -s flag in Makefile

[Riku Voipio]

On Thu, Mar 26, 2009 at 08:50:57PM +0200, Avi Kivity wrote:
> Anthony Liguori wrote:
>> Does anyone object to this?  I know Debian also carries a similar patch.

> If people complain, we can add a ./configure flag for it.  But I doubt  
> anyone will notice; the floppy days are gone.

No need to, you can already pass --install="install -s" to qemu's configure.

-- 
"rm -rf" only sounds scary if you don't have backups

Re: [Qemu-devel] [PATCH] Remove -s flag in Makefile

[Johannes Schindelin]

Hi,

On Wed, 25 Mar 2009, Glauber Costa wrote:

> install -s is pretty and cute, but makes the binary get into the build
> directory without any debug information.

Is it really the build directory?

> This is bad, and makes glommer sad. Also, tools that attempt to extract 
> that data, like fedora debuginfo extractor, will fail.

Is it not common standard to install stripped binaries?  If it is not, 
where is it documented?

Thanks,
Dscho

Re: [Qemu-devel] [PATCH] Remove -s flag in Makefile

[M. Warner Losh]

In message: <alpine.DEB.1.00.0903252106570.26370@intel-tinevez-2-302>
            Johannes Schindelin <Johannes.Schindelin@gmx.de> writes:
: Hi,
: 
: On Wed, 25 Mar 2009, Glauber Costa wrote:
: 
: > install -s is pretty and cute, but makes the binary get into the build
: > directory without any debug information.
: 
: Is it really the build directory?

I wondered about the patch...

: > This is bad, and makes glommer sad. Also, tools that attempt to extract 
: > that data, like fedora debuginfo extractor, will fail.
: 
: Is it not common standard to install stripped binaries?  If it is not, 
: where is it documented?

It is standard.

Warner

Re: [Qemu-devel] [PATCH] Remove -s flag in Makefile

[Anthony Liguori]

M. Warner Losh wrote:
> In message: <alpine.DEB.1.00.0903252106570.26370@intel-tinevez-2-302>
>             Johannes Schindelin <Johannes.Schindelin@gmx.de> writes:
> : Hi,
> : 
> : On Wed, 25 Mar 2009, Glauber Costa wrote:
> : 
> : > install -s is pretty and cute, but makes the binary get into the build
> : > directory without any debug information.
> : 
> : Is it really the build directory?
>
> I wondered about the patch...
>
> : > This is bad, and makes glommer sad. Also, tools that attempt to extract 
> : > that data, like fedora debuginfo extractor, will fail.
> : 
> : Is it not common standard to install stripped binaries?  If it is not, 
> : where is it documented?
>
> It is standard.
>   

Does anyone know what the various ports packages do?  If everyone is 
modifying the Makefile to not strip by default (and they are stripping 
on install), then I think we ought to do the same.  If you're installing 
from source, chances are you'd prefer the executable to have debug symbols.

Regards,

Anthony Liguori

> Warner
>
>
>   

Re: [Qemu-devel] [PATCH] Remove -s flag in Makefile

[Jan Marten Simons]

Am Mittwoch, 25. März 2009 21:45:12 schrieb Anthony Liguori:
> M. Warner Losh wrote:
> >             Johannes Schindelin <Johannes.Schindelin@gmx.de> writes:
> > : Is it not common standard to install stripped binaries?  If it is not,
> > : where is it documented?
> >
> > It is standard.
>
> Does anyone know what the various ports packages do?  If everyone is
> modifying the Makefile to not strip by default (and they are stripping
> on install), then I think we ought to do the same.  If you're installing
> from source, chances are you'd prefer the executable to have debug symbols.

Gentoo's portage per default strips all binaries on install step unless you 
specify FEATURES="nostrip". So there the package management system (=portage) 
is in charge of stripping the binaries. Portage even warns about pre-stiped 
binaries.

I think the default should be to let the binaries unstripped unless the user 
explicitly wants to do this during install.

With regards,
Jan