Skip to contentSkip to conversation

oe-pkgdata-util: avoid glibc-locale being mapped out by glibc packages

PATCH RESEND In review [email protected] 2 participants 3 messages Jul 31, 2025 – Aug 5, 2025

lore
Share & tools
lore
#0 lore raw

PATCH RESEND 1/1 oe-pkgdata-util: avoid glibc-locale being mapped out by glibc packages

From: Chen Qi <[email protected]>

oe-pkgdata-util's glob command will output different results depending on whether glibc-locale is built or not. This in turn results in do_rootfs error when multilib is enabled.

The error message is like below:

No match for argument: lib32-glibc-locale-en-gb
Error: Unable to find a match: lib32-glibc-locale-en-gb

Here are the steps to reproduce the issue:
1. Enable multilib and install a lib32 application by adding the following lines to local.conf:
require conf/multilib.conf
MULTILIBS ?= "multilib:lib32" DEFAULTTUNE:virtclass-multilib-lib32 ?= "core2-32" IMAGE_INSTALL:append = " lib32-sysstat"
2. bitbake lib32-glibc-locale
Note that people usually don't deliberately build lib32-glibc-locale, but they will do world build. And when so do so, the following step fails.
3. bitbake core-image-full-cmdline

This above steps are how to reproduce the issue in oe-core. There are three steps there. So it gives people the impression that it's not easy to reproduce the issue. But in fact, with some image, it could be reproduced by just doing `bitbake lib32-<image>', as long as that <image> brings glibc-locale into its build dependency but not runtime dependency.

The problem happened because do_rootfs will first form its own repo by finding dependencies. glibc-locale will be brought in because of IMAGE_LINGUAS. But lib32-glibc-locale will not be brought in. And in install_complementary, oe-pkgdata-util will use its glob command to find all complememtary packages for IMAGE_LINGUAS, such as "*-locale-en-gb". Then we get the above error.

glibc-locale has its own recipe and which to install into an image is determined by IMAGE_LINGUAS. So let's not map any glibc packages out to glibc-locale in oe-pkgdata-util.

---
 scripts/oe-pkgdata-util | 4 ++++
 1 file changed, 4 insertions(+)
scripts/oe-pkgdata-util +4 -0
index 44ae40549a..9fb7952c58 100755
@@ -136,6 +136,10 @@ def glob(args):
                         logger.debug("%s is not a valid package!" % (pkg))
                         break
 
+                # we don't want glibc-locale to be mapped out by any glibc package, filter it out
+                if "glibc-locale" in mappedpkg:
+                    mappedpkg = ""
+
                 if mappedpkg:
                     logger.debug("%s (%s) -> %s" % (pkg, g, mappedpkg))
                     mappedpkgs.add(mappedpkg)
Signature · 1 line
2.34.1
#1 lore raw
Quoted 49 lines · Chen Qi via lists.openembedded.org

From: Chen Qi<[email protected]>

oe-pkgdata-util's glob command will output different results depending on whether glibc-locale is built or not. This in turn results in do_rootfs error when multilib is enabled.

The error message is like below:

No match for argument: lib32-glibc-locale-en-gb
Error: Unable to find a match: lib32-glibc-locale-en-gb

Here are the steps to reproduce the issue:
1. Enable multilib and install a lib32 application by adding the following lines to local.conf:
require conf/multilib.conf
MULTILIBS ?= "multilib:lib32" DEFAULTTUNE:virtclass-multilib-lib32 ?= "core2-32" IMAGE_INSTALL:append = " lib32-sysstat"
2. bitbake lib32-glibc-locale
Note that people usually don't deliberately build lib32-glibc-locale, but they will do world build. And when so do so, the following step fails.
3. bitbake core-image-full-cmdline

This above steps are how to reproduce the issue in oe-core. There are three steps there. So it gives people the impression that it's not easy to reproduce the issue. But in fact, with some image, it could be reproduced by just doing `bitbake lib32-<image>', as long as that <image> brings glibc-locale into its build dependency but not runtime dependency.

The problem happened because do_rootfs will first form its own repo by finding dependencies. glibc-locale will be brought in because of IMAGE_LINGUAS. But lib32-glibc-locale will not be brought in. And in install_complementary, oe-pkgdata-util will use its glob command to find all complememtary packages for IMAGE_LINGUAS, such as "*-locale-en-gb". Then we get the above error.

glibc-locale has its own recipe and which to install into an image is determined by IMAGE_LINGUAS. So let's not map any glibc packages out to glibc-locale in oe-pkgdata-util.

---
  scripts/oe-pkgdata-util | 4 ++++
  1 file changed, 4 insertions(+)
scripts/oe-pkgdata-util +0 -0
index 44ae40549a..9fb7952c58 100755

scripts/oe-pkgdata-util def glob(args): · line 142

Quoted 7 lines
scripts/oe-pkgdata-util +4 -0
@@ -136,6 +136,10 @@ def glob(args):
                          logger.debug("%s is not a valid package!" % (pkg))
                          break
 
+                # we don't want glibc-locale to be mapped out by any glibc package, filter it out
+                if "glibc-locale" in mappedpkg:
+                    mappedpkg = ""
+

Hi Qi,

During a brief discussion in the patch review meeting,
people didn't like the idea of changing a generic script to deal with a specific recipe problem even if it's glibc and locales.

Can you review/re-explain the discussion from:
https://lore.kernel.org/openembedded-core/PH0PR11MB5611F2E7A0B97861D193A60AED132@PH0PR11MB5611.namprd11.prod.outlook.com/
in a follow-up email  and discuss with co-workers and the list to try to come up with an alternative approach.

I can eventually try to help brainstorm ideas once I've dug out from under this mountain of email that accumulated while I was on holiday.

../Randy

Quoted 12 lines
scripts/oe-pkgdata-util +0 -1
                  if mappedpkg:
                      logger.debug("%s (%s) -> %s" % (pkg, g, mappedpkg))
                      mappedpkgs.add(mappedpkg)
 
-=-=-=-=-=-=-=-=-=-=-=-

Links: You receive all messages sent to this group.
View/Reply Online (#221175):https://lists.openembedded.org/g/openembedded-core/message/221175 Mute This Topic:https://lists.openembedded.org/mt/114463019/3616765
Group Owner:[email protected] Unsubscribe:https://lists.openembedded.org/g/openembedded-core/unsub [[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-

Signature · 2 lines
# Randy MacLeod
# Wind River Linux
author
#2 lore raw

Got it. I’ll revisit this issue and come up with some alternative solution. Any suggestion is welcome.

Regards,
Qi

From: MacLeod, Randy <[email protected]>
Sent: Tuesday, August 5, 2025 7:25 AM
To: Chen, Qi <[email protected]>; [email protected]
Subject: Re: [OE-core][RESEND][PATCH] oe-pkgdata-util: avoid glibc-locale being mapped out by glibc packages

On 2025-07-31 3:39 a.m., Chen Qi via lists.openembedded.org wrote:

From: Chen Qi <[email protected]><mailto:[email protected]>

oe-pkgdata-util's glob command will output different results depending

on whether glibc-locale is built or not. This in turn results in do_rootfs

error when multilib is enabled.

The error message is like below:

No match for argument: lib32-glibc-locale-en-gb

Error: Unable to find a match: lib32-glibc-locale-en-gb

Here are the steps to reproduce the issue:

1. Enable multilib and install a lib32 application by adding the following

lines to local.conf:

require conf/multilib.conf

MULTILIBS ?= "multilib:lib32"

DEFAULTTUNE:virtclass-multilib-lib32 ?= "core2-32"

IMAGE_INSTALL:append = " lib32-sysstat"

2. bitbake lib32-glibc-locale

Note that people usually don't deliberately build lib32-glibc-locale,

but they will do world build. And when so do so, the following step fails.

3. bitbake core-image-full-cmdline

This above steps are how to reproduce the issue in oe-core. There are three

steps there. So it gives people the impression that it's not easy to reproduce

the issue. But in fact, with some image, it could be reproduced by just

doing `bitbake lib32-<image>', as long as that <image> brings glibc-locale

into its build dependency but not runtime dependency.

The problem happened because do_rootfs will first form its own repo by finding

dependencies. glibc-locale will be brought in because of IMAGE_LINGUAS. But

lib32-glibc-locale will not be brought in. And in install_complementary,

oe-pkgdata-util will use its glob command to find all complememtary packages

for IMAGE_LINGUAS, such as "*-locale-en-gb". Then we get the above error.

glibc-locale has its own recipe and which to install into an image is

determined by IMAGE_LINGUAS. So let's not map any glibc packages out

to glibc-locale in oe-pkgdata-util.

---

 scripts/oe-pkgdata-util | 4 ++++
 1 file changed, 4 insertions(+)
scripts/oe-pkgdata-util +0 -0

index 44ae40549a..9fb7952c58 100755

--- a/scripts/oe-pkgdata-util

+++ b/scripts/oe-pkgdata-util

@@ -136,6 +136,10 @@ def glob(args):

logger.debug("%s is not a valid package!" % (pkg))

break

+ # we don't want glibc-locale to be mapped out by any glibc package, filter it out

+ if "glibc-locale" in mappedpkg:

+ mappedpkg = ""

+

Hi Qi,

During a brief discussion in the patch review meeting,
people didn't like the idea of changing a generic script to deal with a specific recipe problem even if it's glibc and locales.

Can you review/re-explain the discussion from:
https://lore.kernel.org/openembedded-core/PH0PR11MB5611F2E7A0B97861D193A60AED132@PH0PR11MB5611.namprd11.prod.outlook.com/
in a follow-up email and discuss with co-workers and the list to try to come up with an alternative approach.

I can eventually try to help brainstorm ideas once I've dug out from under this mountain of email that accumulated while I was on holiday.

../Randy

                 if mappedpkg:

                     logger.debug("%s (%s) -> %s" % (pkg, g, mappedpkg))

                     mappedpkgs.add(mappedpkg)

-=-=-=-=-=-=-=-=-=-=-=-

Links: You receive all messages sent to this group.

View/Reply Online (#221175): https://lists.openembedded.org/g/openembedded-core/message/221175

Mute This Topic: https://lists.openembedded.org/mt/114463019/3616765

Group Owner: [email protected]<mailto:[email protected]>

Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub [[email protected]<mailto:[email protected]>]

-=-=-=-=-=-=-=-=-=-=-=-

--

# Randy MacLeod

# Wind River Linux