[PULL][PATCH 0/1] package.bbclass: fix image_debugfs symlinks

[PULL][PATCH 0/1] package.bbclass: fix image_debugfs symlinks

[William Hauser]

The debugfs does not contain any symlinks, causing gdb to be unable to locate
libraries whose minor versions are abstracted out by using symlinks.
i.e. executables link against libpcap.so/libpcap.so.1, not libpcap.so.1.9.1

This is a result a flawed fstat meant to stop broken symlinks from being copied
to the debugfs. My suggested change removes this fstat in favor of an earlier
check for broken symlinks that is already performed. More details in the commit
message.

This change was tested by examining libpcap's build artifacts.

debugfs/usr/lib/.debug/ contents before:
libpcap.so.1.9.1

debugfs/usr/lib/.debug/ contents after:
libpcap.so -> libpcap.so.1
libpcap.so.1 -> libpcap.so.1.9.1
libpcap.so.1.9.1

William Hauser (1):
  package.bbclass: fix image_debugfs symlinks

 meta/classes/package.bbclass | 8 --------
 1 file changed, 8 deletions(-)

-- 
2.15.1

[PULL][PATCH 1/1] package.bbclass: fix image_debugfs symlinks

[William Hauser]

Allow symlinks to be made in the debugfs by removing an incorrect check that
would always prevent symlinks from being transferred over.

The comment in this code indicates that the code is meant to ensure symlinks
are only copied to the debugfs if their target has already been copied, but
instead checks for the existence of the symlink that has yet to be created.
As a result this check always fails.

This check is safe to get rid of since we will only attempt to copy a
symlink to the debugfs if it's original target exists and is an ELF. It
is, however, still possible for a dead symlink to end up in the debugfs
if its original target was stripped.

Signed-off-by: William Hauser <william.hauser@meraki.com>
---
 meta/classes/package.bbclass | 8 --------
 1 file changed, 8 deletions(-)

diff --git a/meta/classes/package.bbclass b/meta/classes/package.bbclass
index e6236c0bb2..47f3df2e13 100644
--- a/meta/classes/package.bbclass
+++ b/meta/classes/package.bbclass
@@ -1258,14 +1258,6 @@ python split_and_strip_files () {
             src = file[len(dvar):]
             dest = debuglibdir + os.path.dirname(src) + debugdir + "/" + os.path.basename(src) + debugappend
             fpath = dvar + dest
-            # Skip it if the target doesn't exist
-            try:
-                s = os.stat(fpath)
-            except OSError as e:
-                (err, strerror) = e.args
-                if err != errno.ENOENT:
-                    raise
-                continue
 
             ltarget = symlinks[file]
             lpath = os.path.dirname(ltarget)
-- 
2.15.1

Re: [poky] [PULL][PATCH 1/1] package.bbclass: fix image_debugfs symlinks

[Randy MacLeod]

[-- Attachment #1: Type: text/plain, Size: 3461 bytes --]

On 2020-10-08 12:59 p.m., William Hauser via lists.yoctoproject.org wrote:
> The comment in this code indicates that the code is meant to ensure symlinks
> are only copied to the debugfs if their target has already been copied, but
> instead checks for the existence of the symlink that has yet to be created.
> As a result this check always fails.
>
> This check is safe to get rid of since we will only attempt to copy a
> symlink to the debugfs if it's original target exists and is an ELF. It
> is, however, still possible for a dead symlink to end up in the debugfs
> if its original target was stripped.
>
> Signed-off-by: William Hauser <william.hauser@meraki.com>
> ---
>   meta/classes/package.bbclass | 8 --------
>   1 file changed, 8 deletions(-)
>
> diff --git a/meta/classes/package.bbclass b/meta/classes/package.bbclass
> index e6236c0bb2..47f3df2e13 100644
> --- a/meta/classes/package.bbclass
> +++ b/meta/classes/package.bbclass
> @@ -1258,14 +1258,6 @@ python split_and_strip_files () {
>               src = file[len(dvar):]
>               dest = debuglibdir + os.path.dirname(src) + debugdir + "/" + os.path.basename(src) + debugappend
>               fpath = dvar + dest
> -            # Skip it if the target doesn't exist
> -            try:
> -                s = os.stat(fpath)
> -            except OSError as e:
> -                (err, strerror) = e.args
> -                if err != errno.ENOENT:
> -                    raise
> -                continue
>   
>               ltarget = symlinks[file]
>               lpath = os.path.dirname(ltarget)


Hi William,

I think this patch/email got lost because it was sent to the wrong list and
we've been in the last stretch of the gatesgarth release. Could you please
re-send to oe-core if it wasn't merged?

See the README.poky below.

Thanks,

../Randy

Contribution Guidelines
=======================

A guide to submitting patches to OpenEmbedded is available at:

http://www.openembedded.org/wiki/How_to_submit_a_patch_to_OpenEmbedded

There is good documentation on how to write/format patches at:

https://www.openembedded.org/wiki/Commit_Patch_Message_Guidelines


Where to Send Patches
=====================

As Poky is an integration repository (built using a tool called 
combo-layer),
patches against the various components should be sent to their respective
upstreams:

bitbake:
     Git repository: http://git.openembedded.org/bitbake/
     Mailing list: bitbake-devel@lists.openembedded.org

documentation:
     Git repository: http://git.yoctoproject.org/cgit/cgit.cgi/yocto-docs/
     Mailing list: docs@lists.yoctoproject.org

meta-poky, meta-yocto-bsp:
     Git repository: 
http://git.yoctoproject.org/cgit/cgit.cgi/meta-yocto(-bsp)
     Mailing list: poky@lists.yoctoproject.org

Everything else should be sent to the OpenEmbedded Core mailing list.  If in
doubt, check the oe-core git repository for the content you intend to 
modify.
Before sending, be sure the patches apply cleanly to the current oe-core git
repository.

     Git repository: http://git.openembedded.org/openembedded-core/
     Mailing list: openembedded-core@lists.openembedded.org

Note: The scripts directory should be treated with extra care as it is a 
mix of
oe-core and poky-specific files from meta-poky.

>
> 
>

-- 
# Randy MacLeod
# Wind River Linux


[-- Attachment #2: Type: text/html, Size: 5841 bytes --]

Re: [poky] [PULL][PATCH 1/1] package.bbclass: fix image_debugfs symlinks

[William Hauser]

[-- Attachment #1: Type: text/plain, Size: 3719 bytes --]

Sent to openembedded-core@lists.openembedded.org. Thanks, Randy.

On Wed, Nov 4, 2020 at 2:18 PM Randy MacLeod <randy.macleod@windriver.com>
wrote:

> On 2020-10-08 12:59 p.m., William Hauser via lists.yoctoproject.org wrote:
>
> The comment in this code indicates that the code is meant to ensure symlinks
> are only copied to the debugfs if their target has already been copied, but
> instead checks for the existence of the symlink that has yet to be created.
> As a result this check always fails.
>
> This check is safe to get rid of since we will only attempt to copy a
> symlink to the debugfs if it's original target exists and is an ELF. It
> is, however, still possible for a dead symlink to end up in the debugfs
> if its original target was stripped.
>
> Signed-off-by: William Hauser <william.hauser@meraki.com> <william.hauser@meraki.com>
> ---
>  meta/classes/package.bbclass | 8 --------
>  1 file changed, 8 deletions(-)
>
> diff --git a/meta/classes/package.bbclass b/meta/classes/package.bbclass
> index e6236c0bb2..47f3df2e13 100644
> --- a/meta/classes/package.bbclass
> +++ b/meta/classes/package.bbclass
> @@ -1258,14 +1258,6 @@ python split_and_strip_files () {
>              src = file[len(dvar):]
>              dest = debuglibdir + os.path.dirname(src) + debugdir + "/" + os.path.basename(src) + debugappend
>              fpath = dvar + dest
> -            # Skip it if the target doesn't exist
> -            try:
> -                s = os.stat(fpath)
> -            except OSError as e:
> -                (err, strerror) = e.args
> -                if err != errno.ENOENT:
> -                    raise
> -                continue
>
>              ltarget = symlinks[file]
>              lpath = os.path.dirname(ltarget)
>
>
> Hi William,
>
> I think this patch/email got lost because it was sent to the wrong list and
> we've been in the last stretch of the gatesgarth release. Could you please
> re-send to oe-core if it wasn't merged?
>
> See the README.poky below.
>
> Thanks,
>
> ../Randy
>
> Contribution Guidelines
> =======================
>
> A guide to submitting patches to OpenEmbedded is available at:
>
> http://www.openembedded.org/wiki/How_to_submit_a_patch_to_OpenEmbedded
>
> There is good documentation on how to write/format patches at:
>
> https://www.openembedded.org/wiki/Commit_Patch_Message_Guidelines
>
>
> Where to Send Patches
> =====================
>
> As Poky is an integration repository (built using a tool called
> combo-layer),
> patches against the various components should be sent to their respective
> upstreams:
>
> bitbake:
>     Git repository: http://git.openembedded.org/bitbake/
>     Mailing list: bitbake-devel@lists.openembedded.org
>
> documentation:
>     Git repository: http://git.yoctoproject.org/cgit/cgit.cgi/yocto-docs/
>     Mailing list: docs@lists.yoctoproject.org
>
> meta-poky, meta-yocto-bsp:
>     Git repository:
> http://git.yoctoproject.org/cgit/cgit.cgi/meta-yocto(-bsp)
>     Mailing list: poky@lists.yoctoproject.org
>
> Everything else should be sent to the OpenEmbedded Core mailing list.  If
> in
> doubt, check the oe-core git repository for the content you intend to
> modify.
> Before sending, be sure the patches apply cleanly to the current oe-core
> git
> repository.
>
>     Git repository: http://git.openembedded.org/openembedded-core/
>     Mailing list: openembedded-core@lists.openembedded.org
>
> Note: The scripts directory should be treated with extra care as it is a
> mix of
> oe-core and poky-specific files from meta-poky.
>
>
> 
>
>
>
> --
> # Randy MacLeod
> # Wind River Linux
>
>

[-- Attachment #2: Type: text/html, Size: 5728 bytes --]