Skip to contentSkip to conversation

binutils: Pregenerate autoconf files

PATCH In review Khem Raj 3 participants 6 messages Mar 25, 2020 – Mar 27, 2020

lore
Share & tools
lore

Khem Raj

author
#0 lore raw

PATCH 1/1 binutils: Pregenerate autoconf files

libtool patch will result in configure file regeneration, instead of doing that at build time, do it in patch itself, this avoids running autoconf before configure step.

Since binutils needs specific version of autoconf ( which is 2.69 ) this will break on systems using newer or older verisons of autoconf in current state.

---
 meta/recipes-devtools/binutils/binutils.inc   |     8 +-
 .../binutils/0007-Use-libtool-2.4.patch       | 26583 ++++++++++++----
 2 files changed, 20352 insertions(+), 6239 deletions(-)
meta/recipes-devtools/binutils/binutils.inc +1 -7
index b0937e62fc..53f46bb611 100644
@@ -114,13 +114,7 @@ export CC_FOR_BUILD = "LD_LIBRARY_PATH= ${BUILD_CC}"
 MULTIARCH := "${@bb.utils.contains("DISTRO_FEATURES", "multiarch", "yes", "no", d)}"
 do_configure[vardeps] += "MULTIARCH"
 do_configure () {
-	(for d in . bfd binutils gas gold gprof ld libctf opcodes; do
-		cd ${S}/$d
-		autoconf
-		rm -rf autom4te.cache
-	done
-	cd ${S}
-	gnu-configize)
+	(cd ${S} && gnu-configize)
 
 	oe_runconf
 #
meta/recipes-devtools/binutils/binutils/0007-Use-libtool-2.4.patch +20351 -6232

2,986 lines not on this page · raw message

The rest of this diff is not shown. View the raw message.

Signature · 1 line
2.26.0

Part of the diff is not shown here. View the raw message.

#1 lore raw
Quoted 8 lines · Khem Raj

libtool patch will result in configure file regeneration, instead of doing that at build time, do it in patch itself, this avoids running autoconf before configure step.

Since binutils needs specific version of autoconf ( which is 2.69 ) this will break on systems using newer or older verisons of autoconf in current state.

Quoted 7 lines
---
 meta/recipes-devtools/binutils/binutils.inc   |     8 +-
 .../binutils/0007-Use-libtool-2.4.patch       | 26583 ++++++++++++
----
 2 files changed, 20352 insertions(+), 6239 deletions(-)

Whilst I appreciate the intent here, our policy is to autoreconf most things in general. This allows us to more easily support newer architectures and platforms.

There is a significant build speed benefit from not autoreconf'ing things but where do we draw the line?

I'm also worried about patches which touch both configure and configure.ac since the timestamp changes can cause things to autoreconf even when we're trying to avoid that. As such this is actually quite a risky change given past bad experiences :(

I'm not completely against it but I am worried.

Cheers,

Richard

#2 lore raw
Quoted 1 line · Richard Purdie

On Wed, 2020-03-25 at 16:16 -0700, Khem Raj wrote:

...

Quoted 6 lines

I'm also worried about patches which touch both configure and configure.ac since the timestamp changes can cause things to autoreconf even when we're trying to avoid that. As such this is actually quite a risky change given past bad experiences :(

I'm not completely against it but I am worried.

+1

I also prefer to generate them. The patches are easier to upgrade and allow for easier fixes, as mentioned.

Signature · 3 lines
Otavio Salvador                             O.S. Systems
http://www.ossystems.com.br        http://code.ossystems.com.br
Mobile: +55 (53) 9 9981-7854          Mobile: +1 (347) 903-9750
author
#3 lore raw
In reply to Richard Purdie
Quoted 18 lines · Richard Purdie
Quoted 8 lines · Khem Raj

libtool patch will result in configure file regeneration, instead of doing that at build time, do it in patch itself, this avoids running autoconf before configure step.

Since binutils needs specific version of autoconf ( which is 2.69 ) this will break on systems using newer or older verisons of autoconf in current state.

Quoted 7 lines
---
 meta/recipes-devtools/binutils/binutils.inc   |     8 +-
 .../binutils/0007-Use-libtool-2.4.patch       | 26583 ++++++++++++
----
 2 files changed, 20352 insertions(+), 6239 deletions(-)
Quoted 7 lines

Whilst I appreciate the intent here, our policy is to autoreconf most things in general. This allows us to more easily support newer architectures and platforms.

There is a significant build speed benefit from not autoreconf'ing things but where do we draw the line?

in general this is fine but binutils, gcc , glibc can not be treated in general category
since they have dependencies on specific versions of autotools unfortunately, it currently
works for binutils because our version of autoconf matches with what binutils expects
as of now, but this will skew if this changes in future. In nutshell, the auto-fu in these
packages is quite involved and has hard dependencies on specific versions of tools
needed to reconfig them.

Quoted 4 lines

I'm also worried about patches which touch both configure and configure.ac since the timestamp changes can cause things to autoreconf even when we're trying to avoid that. As such this is actually quite a risky change given past bad experiences :(

We were not fully reconfiguring binutils even now, because of other autotool sversion mimatches
only autoconf was being run which does not change the case if configure was say regenerated
as you say.

Quoted 3 lines

I'm not completely against it but I am worried.

perhaps addresses some of your concerns.

Quoted 4 lines

Cheers,

Richard

author
#4 lore raw
Quoted 53 lines · Khem Raj
Quoted 18 lines · Richard Purdie
Quoted 8 lines · Khem Raj

libtool patch will result in configure file regeneration, instead of doing that at build time, do it in patch itself, this avoids running autoconf before configure step.

Since binutils needs specific version of autoconf ( which is 2.69 ) this will break on systems using newer or older verisons of autoconf in current state.

Quoted 7 lines
---
 meta/recipes-devtools/binutils/binutils.inc   |     8 +-
 .../binutils/0007-Use-libtool-2.4.patch       | 26583 ++++++++++++
----
 2 files changed, 20352 insertions(+), 6239 deletions(-)
Quoted 7 lines

Whilst I appreciate the intent here, our policy is to autoreconf most things in general. This allows us to more easily support newer architectures and platforms.

There is a significant build speed benefit from not autoreconf'ing things but where do we draw the line?

in general this is fine but binutils, gcc , glibc can not be treated in general category
since they have dependencies on specific versions of autotools unfortunately, it currently
works for binutils because our version of autoconf matches with what binutils expects
as of now, but this will skew if this changes in future. In nutshell, the auto-fu in these
packages is quite involved and has hard dependencies on specific versions of tools
needed to reconfig them.

Quoted 4 lines

I'm also worried about patches which touch both configure and configure.ac since the timestamp changes can cause things to autoreconf even when we're trying to avoid that. As such this is actually quite a risky change given past bad experiences :(

We were not fully reconfiguring binutils even now, because of other autotool sversion mimatches
only autoconf was being run which does not change the case if configure was say regenerated
as you say.

Quoted 1 line
Quoted 4 lines
Quoted 2 lines

I'm not completely against it but I am worried.

perhaps addresses some of your concerns.

after some IRC discussions, I think it better for us to drop this change.

Quoted 5 lines
Quoted 5 lines

Cheers,

Richard

author
#5 lore raw
Quoted 55 lines · Khem Raj
Quoted 53 lines · Khem Raj
Quoted 18 lines · Richard Purdie
Quoted 8 lines · Khem Raj

libtool patch will result in configure file regeneration, instead of doing that at build time, do it in patch itself, this avoids running autoconf before configure step.

Since binutils needs specific version of autoconf ( which is 2.69 ) this will break on systems using newer or older verisons of autoconf in current state.

Quoted 7 lines
---
 meta/recipes-devtools/binutils/binutils.inc   |     8 +-
 .../binutils/0007-Use-libtool-2.4.patch       | 26583 ++++++++++++
----
 2 files changed, 20352 insertions(+), 6239 deletions(-)
Quoted 7 lines

Whilst I appreciate the intent here, our policy is to autoreconf most things in general. This allows us to more easily support newer architectures and platforms.

There is a significant build speed benefit from not autoreconf'ing things but where do we draw the line?

in general this is fine but binutils, gcc , glibc can not be treated in general category
since they have dependencies on specific versions of autotools unfortunately, it currently
works for binutils because our version of autoconf matches with what binutils expects
as of now, but this will skew if this changes in future. In nutshell, the auto-fu in these
packages is quite involved and has hard dependencies on specific versions of tools
needed to reconfig them.

Quoted 4 lines

I'm also worried about patches which touch both configure and configure.ac since the timestamp changes can cause things to autoreconf even when we're trying to avoid that. As such this is actually quite a risky change given past bad experiences :(

We were not fully reconfiguring binutils even now, because of other autotool sversion mimatches
only autoconf was being run which does not change the case if configure was say regenerated
as you say.

Quoted 1 line
Quoted 7 lines
Quoted 4 lines
Quoted 2 lines

I'm not completely against it but I am worried.

perhaps addresses some of your concerns.

after some IRC discussions, I think it better for us to drop this change.

I think IRC discussions confused me a bit, This patch will improve the situation for binutils since
we dont have to run Autoconf which we are now doing on few selected directories where configure.ac has
been changed due to our patches.

This patch ensures that all configure scripts are up to date and do not require to be regenerated during do_configure
step like common autotooled recipes.

Quoted 5 lines
Quoted 5 lines
Quoted 5 lines

Cheers,

Richard