Skip to contentSkip to conversation

toolchain: control GNU_TARGET_NAME vendor part

PATCH New Noam Camus 3 participants 8 messages Feb 23, 2014 – Mar 4, 2014

lore
Share & tools
lore

Noam Camus

author
#0 lore raw

PATCH 1/1 toolchain: control GNU_TARGET_NAME vendor part

When the toolchain is custom we can denote that by setting in the triplet the vendor part. This is done through configution, where the default is buildroot.

---
 package/Makefile.in                     |    4 +++-
 toolchain/toolchain-buildroot/Config.in |    4 ++++
 2 files changed, 7 insertions(+), 1 deletions(-)
package/Makefile.in +3 -1
@@ -20,8 +20,10 @@ endif
 MAKE1:=$(HOSTMAKE) -j1
 MAKE:=$(HOSTMAKE) $(if $(PARALLEL_JOBS),-j$(PARALLEL_JOBS))
 
+VENDOR:=$(call qstrip,$(BR2_TOOLCHAIN_BUILDROOT_VENDOR))
+
 # Compute GNU_TARGET_NAME
-GNU_TARGET_NAME=$(ARCH)-buildroot-$(TARGET_OS)-$(LIBC)$(ABI)
+GNU_TARGET_NAME=$(ARCH)-$(VENDOR)-$(TARGET_OS)-$(LIBC)$(ABI)
 
 # Blackfin FLAT needs uclinux
 ifeq ($(BR2_bfin)$(BR2_BINFMT_FLAT),yy)
toolchain/toolchain-buildroot/Config.in +4 -0
index cd88889..d1760dd 100644
@@ -69,6 +69,10 @@ config BR2_TOOLCHAIN_BUILDROOT_LIBC
 	default "glibc"  if BR2_TOOLCHAIN_BUILDROOT_EGLIBC
 	default "glibc"  if BR2_TOOLCHAIN_BUILDROOT_GLIBC
 
+config BR2_TOOLCHAIN_BUILDROOT_VENDOR
+	string "custom buildroot toolchain vendor name"
+	default "buildroot"
+
 source "package/uclibc/Config.in"
 
 source "package/binutils/Config.in.host"

--
1.7.1

#1 lore raw

PATCH 1/1 toolchain: control GNU_TARGET_NAME vendor part

Dear Noam Camus,

Quoted 4 lines · Noam Camus

When the toolchain is custom we can denote that by setting in the triplet the vendor part. This is done through configution, where the default is buildroot.

Is there a compelling reason to allow the configuration of the vendor part of the tuple? To me, it makes sense that all Buildroot toolchain have the "buildroot" vendor, so that they can be identified easily.

Also, if by mistake the vendor part of the tuple is defined by the user as "unknown", they he might get compilation issues when building with target architecture == host architecture.

See 11017f081fc5b034e680d89eaea729c19f450e01 ('pkg-infra: make sure cross compiling is enabled when host == target').

Thomas

Signature · 3 lines
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux, Kernel and Android engineering
http://free-electrons.com
author
#2 lore raw

PATCH 1/1 toolchain: control GNU_TARGET_NAME vendor part

Hi Thomas,

Here (EZchip) we change toolchain elements such binutils/gdb kernel (headers) and gcc.
We are using override feature of buildroot to create our custom buildroot toolchain.
One of the GCC changes we carry is that we assume dedicated vendor (ezchip) to control several GCC defaults and enable extra configurations within GCC.

About "unknown" I guess that it can be blocked by checking the value and in some cases cause intentional make error.

I hope this makes sense to you.

Noam

-----Original Message-----
From: Thomas Petazzoni [mailto:thomas.petazzoni at free-electrons.com]
Sent: Sunday, February 23, 2014 11:47 AM
To: Noam Camus
Cc: buildroot at busybox.net
Subject: Re: [Buildroot] [PATCH] toolchain: control GNU_TARGET_NAME vendor part

Dear Noam Camus,

Quoted 4 lines · Noam Camus

When the toolchain is custom we can denote that by setting in the triplet the vendor part. This is done through configution, where the default is buildroot.

Is there a compelling reason to allow the configuration of the vendor part of the tuple? To me, it makes sense that all Buildroot toolchain have the "buildroot" vendor, so that they can be identified easily.

Also, if by mistake the vendor part of the tuple is defined by the user as "unknown", they he might get compilation issues when building with target architecture == host architecture.

See 11017f081fc5b034e680d89eaea729c19f450e01 ('pkg-infra: make sure cross compiling is enabled when host == target').

Thomas
--
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux, Kernel and Android engineering http://free-electrons.com

#3 lore raw

PATCH 1/1 toolchain: control GNU_TARGET_NAME vendor part

Dear Noam Camus,

(Please do not top post. It would also be nice if you used a mail client that wrapped lines at a reasonable length.)

Quoted 3 lines · Noam Camus

Here (EZchip) we change toolchain elements such binutils/gdb kernel (headers) and gcc.
We are using override feature of buildroot to create our custom buildroot toolchain.
One of the GCC changes we carry is that we assume dedicated vendor (ezchip) to control several GCC defaults and enable extra configurations within GCC.

Hum, ok. Is the vendor part of the tuple meant to be used this way?

Quoted 3 lines

About "unknown" I guess that it can be blocked by checking the value and in some cases cause intentional make error.

I hope this makes sense to you.

Right. I guess we can make this configurable. However, your patch has some issues:

* It only creates the option BR2_TOOLCHAIN_BUILDROOT_VENDOR in the internal toolchain backend. But GNU_TARGET_NAME is also used with the external toolchain backend, which means in the external toolchain case we would end up with a GNU_TARGET_NAME having an empty vendor part of the tuple.

* The new Config.in option lacks an help text, and the help text should explain in detail how this should be used, and in which situations. Also the title of the help text should be just "custom toolchain vendor name".

Thanks!

Thomas

Signature · 3 lines
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux, Kernel and Android engineering
http://free-electrons.com
author
#4 lore raw

PATCH 1/1 toolchain: control GNU_TARGET_NAME vendor part

Quoted 9 lines

Right. I guess we can make this configurable. However, your patch has some issues:

* It only creates the option BR2_TOOLCHAIN_BUILDROOT_VENDOR in the internal toolchain backend. But GNU_TARGET_NAME is also used with the external toolchain backend, which means in the external toolchain case we would end up with a GNU_TARGET_NAME having an empty vendor part of the tuple.

Quoted 4 lines

* The new Config.in option lacks an help text, and the help text should explain in detail how this should be used, and in which situations. Also the title of the help text should be just "custom toolchain vendor name".

Below is the revised patch.
Is it OK?
Is it the right way to bring up the corrections?
====================================================

When the toolchain is custom we can denote that by setting in the triplet the vendor part. This is done through configution, where the default is buildroot.

---
 package/Makefile.in                     |   12 +++++++++++-
 toolchain/toolchain-buildroot/Config.in |   12 ++++++++++++
 2 files changed, 23 insertions(+), 1 deletions(-)
package/Makefile.in +10 -1
@@ -20,8 +20,18 @@ endif
 MAKE1:=$(HOSTMAKE) -j1
 MAKE:=$(HOSTMAKE) $(if $(PARALLEL_JOBS),-j$(PARALLEL_JOBS))
 
+ifeq ($(BR2_TOOLCHAIN_BUILDROOT_VENDOR),)
+VENDOR:=buildroot
+else
+ifeq ($(BR2_TOOLCHAIN_BUILDROOT_VENDOR),unknown)
+$(error You may not provide 'unknown' as vendor to toolchain tuple) 
+else VENDOR:=$(call qstrip,$(BR2_TOOLCHAIN_BUILDROOT_VENDOR))
+endif
+endif
+
 # Compute GNU_TARGET_NAME
-GNU_TARGET_NAME=$(ARCH)-buildroot-$(TARGET_OS)-$(LIBC)$(ABI)
+GNU_TARGET_NAME=$(ARCH)-$(VENDOR)-$(TARGET_OS)-$(LIBC)$(ABI)
 
 # Blackfin FLAT needs uclinux
 ifeq ($(BR2_bfin)$(BR2_BINFMT_FLAT),yy)
toolchain/toolchain-buildroot/Config.in +12 -0
index cd88889..216f308 100644
@@ -69,6 +69,18 @@ config BR2_TOOLCHAIN_BUILDROOT_LIBC
 	default "glibc"  if BR2_TOOLCHAIN_BUILDROOT_EGLIBC
 	default "glibc"  if BR2_TOOLCHAIN_BUILDROOT_GLIBC
 
+config BR2_TOOLCHAIN_BUILDROOT_VENDOR
+	string "custom toolchain vendor name"
+	default "buildroot"
+	help
+	  This option accepts name for vendor field in toolchain tuple.
+	  The toolchain tuple full format is:
+	  cpu-vendor-kernel-os
+	  You may use it in case of custom toolchain.
+	  e.g. if your gcc depend on this field to make decisions.
+	  NOTE: "unknown" is not allowed since it may be confused
+	  with host toolchain.
+
 source "package/uclibc/Config.in"
 
 source "package/binutils/Config.in.host"

--
1.7.1

Thanks
Noam

#5 lore raw

PATCH 1/1 toolchain: control GNU_TARGET_NAME vendor part

Noam, All,

On 2014-02-23 15:41 +0000, Noam Camus spake thusly:
[--SNIP--]

Quoted 3 lines

Below is the revised patch.
Is it OK?
Is it the right way to bring up the corrections?

Please just re-send the patch as you did the first time. Be sure to annotate it as:
[PATCH v2] blabla

You can do that with:

    git send-email --annotate [...]

[--SNIP--]

Quoted 9 lines
toolchain/toolchain-buildroot/Config.in +1 -0
index cd88889..216f308 100644
@@ -69,6 +69,18 @@ config BR2_TOOLCHAIN_BUILDROOT_LIBC
 	default "glibc"  if BR2_TOOLCHAIN_BUILDROOT_EGLIBC
 	default "glibc"  if BR2_TOOLCHAIN_BUILDROOT_GLIBC
 
+config BR2_TOOLCHAIN_BUILDROOT_VENDOR

toolchain/toolchain-buildroot/Config.in config BR2_TOOLCHAIN_BUILDROOT_LIBC · line 82

Quoted 10 lines
toolchain/toolchain-buildroot/Config.in +10 -0
+	string "custom toolchain vendor name"
+	default "buildroot"
+	help
+	  This option accepts name for vendor field in toolchain tuple.
+	  The toolchain tuple full format is:
+	  cpu-vendor-kernel-os
+	  You may use it in case of custom toolchain.
+	  e.g. if your gcc depend on this field to make decisions.
+	  NOTE: "unknown" is not allowed since it may be confused
+	  with host toolchain.

I think Thomas said on IRC that he will have a better wording for the help text. Wait a little bit for his suggestion, before re-sending.

Thanks!

Regards,
Yann E. MORIN.

Signature · 6 lines
.-----------------.--------------------.------------------.--------------------.
|  Yann E. MORIN  | Real-Time Embedded | /"\ ASCII RIBBON | Erics' conspiracy: |
| +33 662 376 056 | Software  Designer | \ / CAMPAIGN     |  ___               |
| +33 223 225 172 `------------.-------:  X  AGAINST      |  \e/  There is no  |
| http://ymorin.is-a-geek.org/ | _/*\_ | / \ HTML MAIL    |   v   conspiracy.  |
'------------------------------^-------^------------------^--------------------'
#6 lore raw

PATCH 1/1 toolchain: control GNU_TARGET_NAME vendor part

In reply to Noam Camus

Dear Noam Camus,

toolchain/toolchain-buildroot/Config.in config BR2_TOOLCHAIN_BUILDROOT_LIBC · line 82

Quoted 11 lines · Noam Camus
toolchain/toolchain-buildroot/Config.in +11 -0
+config BR2_TOOLCHAIN_BUILDROOT_VENDOR
+	string "custom toolchain vendor name"
+	default "buildroot"
+	help
+	  This option accepts name for vendor field in toolchain tuple.
+	  The toolchain tuple full format is:
+	  cpu-vendor-kernel-os
+	  You may use it in case of custom toolchain.
+	  e.g. if your gcc depend on this field to make decisions.
+	  NOTE: "unknown" is not allowed since it may be confused
+	  with host toolchain.

A better wording IMO:

This option allows to customize the "vendor" part of the toolchain tuple, where the toolchain tuple has the form <cpu>-<vendor>-<kernel>-<os>. The default value, "buildroot", is fine for most cases, except in very specific situations where gcc might make different decisions based on the vendor part of the tuple. The value "unknown" is not allowed as the cross-compiling toolchain might then be confused with the native toolchain when the target and host architecture are identical.

If you're not sure, just leave the default "buildroot" value.

Thomas

Signature · 3 lines
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux, Kernel and Android engineering
http://free-electrons.com
author
#7 lore raw

PATCH 1/1 toolchain: control GNU_TARGET_NAME vendor part

Quoted 12 lines

A better wording IMO:

This option allows to customize the "vendor" part of the
toolchain tuple, where the toolchain tuple has the form <cpu>-<vendor>-<kernel>-<os>. The default value, "buildroot", is fine for most cases, except in very specific situations where gcc might make different decisions based on the vendor part of the tuple. The value "unknown" is not allowed as the cross-compiling toolchain might then be confused with the native toolchain when the target and host architecture are identical.

Quoted 1 line

If you're not sure, just leave the default "buildroot" value.

Thanks
It is indeed much better.
I will send PATCH V2
________________________________________
From: Thomas Petazzoni [thomas.petazzoni at free-electrons.com]
Sent: Monday, March 03, 2014 11:44 PM
To: Noam Camus
Cc: buildroot at busybox.net
Subject: Re: [Buildroot] [PATCH] toolchain: control GNU_TARGET_NAME vendor part

Dear Noam Camus,

toolchain/toolchain-buildroot/Config.in config BR2_TOOLCHAIN_BUILDROOT_LIBC · line 82

Quoted 11 lines · Noam Camus
toolchain/toolchain-buildroot/Config.in +11 -0
+config BR2_TOOLCHAIN_BUILDROOT_VENDOR
+     string "custom toolchain vendor name"
+     default "buildroot"
+     help
+       This option accepts name for vendor field in toolchain tuple.
+       The toolchain tuple full format is:
+       cpu-vendor-kernel-os
+       You may use it in case of custom toolchain.
+       e.g. if your gcc depend on this field to make decisions.
+       NOTE: "unknown" is not allowed since it may be confused
+       with host toolchain.

A better wording IMO:

This option allows to customize the "vendor" part of the toolchain tuple, where the toolchain tuple has the form <cpu>-<vendor>-<kernel>-<os>. The default value, "buildroot", is fine for most cases, except in very specific situations where gcc might make different decisions based on the vendor part of the tuple. The value "unknown" is not allowed as the cross-compiling toolchain might then be confused with the native toolchain when the target and host architecture are identical.

If you're not sure, just leave the default "buildroot" value.

Thomas
--
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux, Kernel and Android engineering
http://free-electrons.com