Skip to contentSkip to conversation

package_ipk: handle exception for subprocess command

PATCH In review Andrey Zhizhikin 2 participants 10 messages Mar 28, 2019 – Apr 25, 2019

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 package_ipk: handle exception for subprocess command

When opkg-build command fails to execute, subprocess is returned with exception instead of printing to stderr. This causes the error logging not to be printed out, as the "finally" statement does not contain any bitbake error output.

One example of this behavior is when the package name contains uppercase character, which are rejected by opkg-build, subprocess.check_output would except and no error log would be produced.

This commit catches the exception subprocess.CalledProcessError and produces bb.error output visible to the user.

---
 meta/classes/package_ipk.bbclass | 2 ++
 1 file changed, 2 insertions(+)
meta/classes/package_ipk.bbclass +2 -0
index d1b317b42b..f181f5b4fd 100644
@@ -234,6 +234,8 @@ def ipk_write_pkg(pkg, d):
             ipk_to_sign = "%s/%s_%s_%s.ipk" % (pkgoutdir, pkgname, ipkver, d.getVar('PACKAGE_ARCH'))
             sign_ipk(d, ipk_to_sign)
 
+    except subprocess.CalledProcessError as exc:
+        bb.error("OPKG Build failed: %s" % exc.output)
     finally:
         cleanupcontrol(root)
         bb.utils.unlockfile(lf)
Signature · 1 line
2.17.1
#1 lore raw

Ping.

Quoted 34 lines · Andrey Zhizhikin

When opkg-build command fails to execute, subprocess is returned with exception instead of printing to stderr. This causes the error logging not to be printed out, as the "finally" statement does not contain any bitbake error output.

One example of this behavior is when the package name contains uppercase character, which are rejected by opkg-build, subprocess.check_output would except and no error log would be produced.

This commit catches the exception subprocess.CalledProcessError and produces bb.error output visible to the user.

---
 meta/classes/package_ipk.bbclass | 2 ++
 1 file changed, 2 insertions(+)
meta/classes/package_ipk.bbclass +2 -1
index d1b317b42b..f181f5b4fd 100644
@@ -234,6 +234,8 @@ def ipk_write_pkg(pkg, d):
             ipk_to_sign = "%s/%s_%s_%s.ipk" % (pkgoutdir, pkgname, ipkver, d.getVar('PACKAGE_ARCH'))
             sign_ipk(d, ipk_to_sign)
 
+    except subprocess.CalledProcessError as exc:
+        bb.error("OPKG Build failed: %s" % exc.output)
     finally:
         cleanupcontrol(root)
         bb.utils.unlockfile(lf)
--

2.17.1

#2 lore raw
Quoted 36 lines · Andrey Zhizhikin

Ping.

On Thu, Mar 28, 2019 at 10:47 AM Andrey Zhizhikin <[email protected]

Quoted 33 lines

wrote:
When opkg-build command fails to execute, subprocess is returned with
exception instead of printing to stderr. This causes the error logging
not to be printed out, as the "finally" statement does not contain any
bitbake error output.

One example of this behavior is when the package name contains uppercase
character, which are rejected by opkg-build, subprocess.check_output
would except and no error log would be produced.

This commit catches the exception subprocess.CalledProcessError and produces bb.error output visible to the user.

---
 meta/classes/package_ipk.bbclass | 2 ++
 1 file changed, 2 insertions(+)
meta/classes/package_ipk.bbclass +0 -0
index d1b317b42b..f181f5b4fd 100644
@@ -234,6 +234,8 @@ def ipk_write_pkg(pkg, d):
             ipk_to_sign = "%s/%s_%s_%s.ipk" % (pkgoutdir, pkgname,
 ipkver, d.getVar('PACKAGE_ARCH'))
             sign_ipk(d, ipk_to_sign)
 
Quoted 5 lines
Quoted 5 lines
meta/classes/package_ipk.bbclass +2 -0
+    except subprocess.CalledProcessError as exc:
+        bb.error("OPKG Build failed: %s" % exc.output)
     finally:
         cleanupcontrol(root)

bb.utils.unlockfile(lf)

My main concern is why isn't the raised exception being caught and causing its own error...

This feels like a workaround rather than fixing the underlying problem which I suspect might be in the parallel execution code exception handling.

Cheers,

Richard

#3 lore raw
Quoted 38 lines · Richard Purdie
Quoted 36 lines · Andrey Zhizhikin

Ping.

On Thu, Mar 28, 2019 at 10:47 AM Andrey Zhizhikin <[email protected]

Quoted 33 lines

wrote:
When opkg-build command fails to execute, subprocess is returned with
exception instead of printing to stderr. This causes the error logging
not to be printed out, as the "finally" statement does not contain any
bitbake error output.

One example of this behavior is when the package name contains uppercase
character, which are rejected by opkg-build, subprocess.check_output
would except and no error log would be produced.

This commit catches the exception subprocess.CalledProcessError and produces bb.error output visible to the user.

---
 meta/classes/package_ipk.bbclass | 2 ++
 1 file changed, 2 insertions(+)
meta/classes/package_ipk.bbclass +0 -0
index d1b317b42b..f181f5b4fd 100644
@@ -234,6 +234,8 @@ def ipk_write_pkg(pkg, d):
             ipk_to_sign = "%s/%s_%s_%s.ipk" % (pkgoutdir, pkgname,
 ipkver, d.getVar('PACKAGE_ARCH'))
             sign_ipk(d, ipk_to_sign)
 
Quoted 8 lines
Quoted 5 lines
Quoted 5 lines
meta/classes/package_ipk.bbclass +2 -0
+    except subprocess.CalledProcessError as exc:
+        bb.error("OPKG Build failed: %s" % exc.output)
     finally:
         cleanupcontrol(root)

bb.utils.unlockfile(lf)

My main concern is why isn't the raised exception being caught and causing its own error...

The raised exception is actually caught by a finally: statement below, and the build gracefully terminates. The problem is that finally:
block does not contain any valuable output to inform user what actually happened.
subprocess.check_output() would throw this exception every time the command in sub-process is terminated with the error code, and since we do tell it to dump stderr -> stdout - the error message would be contained in the exception output.
This additional handling of the subprocess.check_output() exception would extract the stdout from the failed process here and just shows to the user the actual output from command processing, so that he is aware what was wrong.

The case where I personally needed it the most is when the package name contained upper and lower case characters, which were rejected by the opkg-build command and until I introduced the handler - I just had an erroneous build failure without any additional information on what went wrong.

Quoted 4 lines

This feels like a workaround rather than fixing the underlying problem which I suspect might be in the parallel execution code exception handling.

The exception from subprocess.check_output() is actually expected and perfectly handled, so there is no problem with that. This patch would just deliver a bit more information in the output for user to react proper.

Quoted 6 lines

Cheers,

Richard

Signature · 2 lines
Regards,
Andrey.
#4 lore raw
Quoted 53 lines · Andrey Zhizhikin
Quoted 42 lines · Richard Purdie
Quoted 41 lines · Andrey Zhizhikin

Ping.

On Thu, Mar 28, 2019 at 10:47 AM Andrey Zhizhikin < [email protected]

Quoted 37 lines

wrote:
When opkg-build command fails to execute, subprocess is returned
with
exception instead of printing to stderr. This causes the error logging
not to be printed out, as the "finally" statement does not contain
any
bitbake error output.

One example of this behavior is when the package name contains uppercase
character, which are rejected by opkg-build, subprocess.check_output
would except and no error log would be produced.

This commit catches the exception subprocess.CalledProcessError and
produces bb.error output visible to the user.

---
 meta/classes/package_ipk.bbclass | 2 ++
 1 file changed, 2 insertions(+)
meta/classes/package_ipk.bbclass +0 -0
index d1b317b42b..f181f5b4fd 100644
@@ -234,6 +234,8 @@ def ipk_write_pkg(pkg, d):
             ipk_to_sign = "%s/%s_%s_%s.ipk" % (pkgoutdir,
 pkgname,
 ipkver, d.getVar('PACKAGE_ARCH'))
             sign_ipk(d, ipk_to_sign)
 
Quoted 8 lines
Quoted 5 lines
Quoted 5 lines
meta/classes/package_ipk.bbclass +2 -0
+    except subprocess.CalledProcessError as exc:
+        bb.error("OPKG Build failed: %s" % exc.output)
     finally:

cleanupcontrol(root)
bb.utils.unlockfile(lf)

My main concern is why isn't the raised exception being caught and causing its own error...

Quoted 4 lines

The raised exception is actually caught by a finally: statement below, and the build gracefully terminates. The problem is that
finally: block does not contain any valuable output to inform user what actually happened.

This isn't how python works. The exception should be "re-raised after the finally clause has been executed" to quote the python manual:
https://docs.python.org/3/tutorial/errors.html#defining-clean-up-actions

Quoted 22 lines

subprocess.check_output() would throw this exception every time the command in sub-process is terminated with the error code, and since we
do tell it to dump stderr -> stdout - the error message would be contained in the exception output.
This additional handling of the subprocess.check_output() exception would extract the stdout from the failed process here and just shows to the user the actual output from command processing, so that he is aware what was wrong.

The case where I personally needed it the most is when the package name contained upper and lower case characters, which were rejected by
the opkg-build command and until I introduced the handler - I just had
an erroneous build failure without any additional information on what went wrong.

Quoted 4 lines

This feels like a workaround rather than fixing the underlying problem
which I suspect might be in the parallel execution code exception handling.

Quoted 5 lines

The exception from subprocess.check_output() is actually expected and
perfectly handled, so there is no problem with that. This patch would just deliver a bit more information in the output for user to react proper.

See my comment above, the finally should not be blocking this exception from being raised...

Cheers,

Richard

#5 lore raw
Quoted 55 lines · On Tue, Apr 16, 2019 at 10:24 AM <[email protected]> wrote:
Quoted 53 lines · Andrey Zhizhikin
Quoted 50 lines · Richard Purdie
Quoted 41 lines · Andrey Zhizhikin

Ping.

On Thu, Mar 28, 2019 at 10:47 AM Andrey Zhizhikin < [email protected]

Quoted 37 lines

wrote:
When opkg-build command fails to execute, subprocess is returned
with
exception instead of printing to stderr. This causes the error logging
not to be printed out, as the "finally" statement does not contain
any
bitbake error output.

One example of this behavior is when the package name contains uppercase
character, which are rejected by opkg-build, subprocess.check_output
would except and no error log would be produced.

This commit catches the exception subprocess.CalledProcessError and
produces bb.error output visible to the user.

---
 meta/classes/package_ipk.bbclass | 2 ++
 1 file changed, 2 insertions(+)
meta/classes/package_ipk.bbclass +0 -0
index d1b317b42b..f181f5b4fd 100644
@@ -234,6 +234,8 @@ def ipk_write_pkg(pkg, d):
             ipk_to_sign = "%s/%s_%s_%s.ipk" % (pkgoutdir,
 pkgname,
 ipkver, d.getVar('PACKAGE_ARCH'))
             sign_ipk(d, ipk_to_sign)
 
Quoted 5 lines
Quoted 5 lines
meta/classes/package_ipk.bbclass +2 -0
+    except subprocess.CalledProcessError as exc:
+        bb.error("OPKG Build failed: %s" % exc.output)
     finally:

cleanupcontrol(root)
bb.utils.unlockfile(lf)

My main concern is why isn't the raised exception being caught and causing its own error...

Quoted 9 lines
Quoted 4 lines

The raised exception is actually caught by a finally: statement below, and the build gracefully terminates. The problem is that
finally: block does not contain any valuable output to inform user what actually happened.

This isn't how python works. The exception should be "re-raised after the finally clause has been executed" to quote the python manual:
https://docs.python.org/3/tutorial/errors.html#defining-clean-up-actions

Sorry, I guess my previous reply was a bit confusing.. I agree, the exception would not be blocked by finally: statement, and this is why the build gracefully shuts down. What the finally: block does not contain is an bb.error() which would provide more information about the source of error return from subprocess.check_output(). In case if this patch is applied - exception would be handled and not propagated further.

Can you please advise whether there would another "raise" statement be needed after bb.error in the patch, so that in addition to the subprocess output user would get an entire callstack (like in the case when subprocess.CalledProcessError was not handled). Currently, with this patch user would receive the build error with the error string output from subprocess.check_output().

Thanks a lot!

Quoted 36 lines
Quoted 22 lines

subprocess.check_output() would throw this exception every time the command in sub-process is terminated with the error code, and since we
do tell it to dump stderr -> stdout - the error message would be contained in the exception output.
This additional handling of the subprocess.check_output() exception would extract the stdout from the failed process here and just shows to the user the actual output from command processing, so that he is aware what was wrong.

The case where I personally needed it the most is when the package name contained upper and lower case characters, which were rejected by
the opkg-build command and until I introduced the handler - I just had
an erroneous build failure without any additional information on what went wrong.

Quoted 4 lines

This feels like a workaround rather than fixing the underlying problem
which I suspect might be in the parallel execution code exception handling.

Quoted 5 lines

The exception from subprocess.check_output() is actually expected and
perfectly handled, so there is no problem with that. This patch would just deliver a bit more information in the output for user to react proper.

See my comment above, the finally should not be blocking this exception from being raised...

Cheers,

Richard

Signature · 2 lines
Regards,
Andrey.
#6 lore raw

Hello Richard,

Quoted 75 lines · Andrey Zhizhikin
Quoted 55 lines · On Tue, Apr 16, 2019 at 10:24 AM <[email protected]> wrote:
Quoted 53 lines · Andrey Zhizhikin
Quoted 50 lines · Richard Purdie
Quoted 41 lines · Andrey Zhizhikin

Ping.

On Thu, Mar 28, 2019 at 10:47 AM Andrey Zhizhikin < [email protected]

Quoted 37 lines

wrote:
When opkg-build command fails to execute, subprocess is returned
with
exception instead of printing to stderr. This causes the error logging
not to be printed out, as the "finally" statement does not contain
any
bitbake error output.

One example of this behavior is when the package name contains uppercase
character, which are rejected by opkg-build, subprocess.check_output
would except and no error log would be produced.

This commit catches the exception subprocess.CalledProcessError and
produces bb.error output visible to the user.

---
 meta/classes/package_ipk.bbclass | 2 ++
 1 file changed, 2 insertions(+)
meta/classes/package_ipk.bbclass +0 -0
index d1b317b42b..f181f5b4fd 100644
@@ -234,6 +234,8 @@ def ipk_write_pkg(pkg, d):
             ipk_to_sign = "%s/%s_%s_%s.ipk" % (pkgoutdir,
 pkgname,
 ipkver, d.getVar('PACKAGE_ARCH'))
             sign_ipk(d, ipk_to_sign)
 
Quoted 5 lines
Quoted 5 lines
meta/classes/package_ipk.bbclass +2 -0
+    except subprocess.CalledProcessError as exc:
+        bb.error("OPKG Build failed: %s" % exc.output)
     finally:

cleanupcontrol(root)
bb.utils.unlockfile(lf)

My main concern is why isn't the raised exception being caught and causing its own error...

Quoted 9 lines
Quoted 4 lines

The raised exception is actually caught by a finally: statement below, and the build gracefully terminates. The problem is that
finally: block does not contain any valuable output to inform user what actually happened.

This isn't how python works. The exception should be "re-raised after the finally clause has been executed" to quote the python manual:
https://docs.python.org/3/tutorial/errors.html#defining-clean-up-actions

Sorry, I guess my previous reply was a bit confusing.. I agree, the exception would not be blocked by finally: statement, and this is why the build gracefully shuts down. What the finally: block does not contain is an bb.error() which would provide more information about the source of error return from subprocess.check_output(). In case if this patch is applied - exception would be handled and not propagated further.

Quoted 8 lines

Can you please advise whether there would another "raise" statement be needed after bb.error in the patch, so that in addition to the subprocess output user would get an entire callstack (like in the case when subprocess.CalledProcessError was not handled). Currently, with this patch user would receive the build error with the error string output from subprocess.check_output().

Thanks a lot!

Can we follow-up on this patch? I'd really appreciate if you can comment on my points here...

Thanks a lot!

Quoted 33 lines
Quoted 33 lines
Quoted 23 lines

subprocess.check_output() would throw this exception every time the command in sub-process is terminated with the error code, and since we
do tell it to dump stderr -> stdout - the error message would be contained in the exception output.
This additional handling of the subprocess.check_output() exception would extract the stdout from the failed process here and just shows to the user the actual output from command processing, so that he is aware what was wrong.

The case where I personally needed it the most is when the package name contained upper and lower case characters, which were rejected by
the opkg-build command and until I introduced the handler - I just had
an erroneous build failure without any additional information on what went wrong.

Quoted 4 lines

This feels like a workaround rather than fixing the underlying problem
which I suspect might be in the parallel execution code exception handling.

Quoted 5 lines

The exception from subprocess.check_output() is actually expected and
perfectly handled, so there is no problem with that. This patch would just deliver a bit more information in the output for user to react proper.

See my comment above, the finally should not be blocking this exception from being raised...

Quoted 8 lines
Quoted 4 lines

Cheers,

Richard

--
Regards,
Andrey.

--
Regards,
Andrey.

#7 lore raw
In reply to Andrey Zhizhikin
Quoted 80 lines · Andrey Zhizhikin
Quoted 64 lines · On Tue, Apr 16, 2019 at 10:24 AM <[email protected]> wrote:
Quoted 57 lines · Andrey Zhizhikin
Quoted 45 lines · Richard Purdie
Quoted 44 lines · Andrey Zhizhikin

Ping.

On Thu, Mar 28, 2019 at 10:47 AM Andrey Zhizhikin < [email protected]

Quoted 40 lines

wrote:
When opkg-build command fails to execute, subprocess is returned
with
exception instead of printing to stderr. This causes the error
logging
not to be printed out, as the "finally" statement does not contain
any
bitbake error output.

One example of this behavior is when the package name contains
uppercase
character, which are rejected by opkg-build, subprocess.check_output
would except and no error log would be produced.

This commit catches the exception
subprocess.CalledProcessError
and
produces bb.error output visible to the user.

---
 meta/classes/package_ipk.bbclass | 2 ++
 1 file changed, 2 insertions(+)
meta/classes/package_ipk.bbclass +0 -0
index d1b317b42b..f181f5b4fd 100644
@@ -234,6 +234,8 @@ def ipk_write_pkg(pkg, d):
             ipk_to_sign = "%s/%s_%s_%s.ipk" % (pkgoutdir,
 pkgname,
 ipkver, d.getVar('PACKAGE_ARCH'))
             sign_ipk(d, ipk_to_sign)
 
Quoted 9 lines
Quoted 5 lines
Quoted 5 lines
meta/classes/package_ipk.bbclass +2 -0
+    except subprocess.CalledProcessError as exc:
+        bb.error("OPKG Build failed: %s" % exc.output)
     finally:

cleanupcontrol(root)
bb.utils.unlockfile(lf)

My main concern is why isn't the raised exception being caught and
causing its own error...

Quoted 5 lines

The raised exception is actually caught by a finally: statement below, and the build gracefully terminates. The problem is that
finally: block does not contain any valuable output to inform user
what actually happened.

Quoted 5 lines

This isn't how python works. The exception should be "re-raised after
the finally clause has been executed" to quote the python manual:
https://docs.python.org/3/tutorial/errors.html#defining-clean-up-actions

Sorry, I guess my previous reply was a bit confusing.. I agree, the exception would not be blocked by finally: statement, and this is why the build gracefully shuts down. What the finally: block does not contain is an bb.error() which would provide more information about the source of error return from subprocess.check_output(). In case if this patch is applied - exception would be handled and not propagated further.

Quoted 8 lines

Can you please advise whether there would another "raise" statement be
needed after bb.error in the patch, so that in addition to the subprocess output user would get an entire callstack (like in the case
when subprocess.CalledProcessError was not handled). Currently, with this patch user would receive the build error with the error string output from subprocess.check_output().

My worry is that we're making a special case fix and for example the other package back ends could have a similar problem (or any other users of multiprocess_launch).

I already think subprocess in python is a bit broken as it should share e.output if its present in an exception. We already special case that in bitbake, the problem here is that our special case code doesn't catch this.

Whilst still ugly, perhaps a better fix might be:

meta/lib/oe/utils.py +6 -1
index a4fd79ccb21..59251810d43 100644
@@ -324,7 +324,12 @@ def multiprocess_launch(target, items, d, extraargs=None):
     if errors:
         msg = ""
         for (e, tb) in errors:
-            msg = msg + str(e) + ": " + str(tb) + "\n"
+            if isinstance(e, subprocess.CalledProcessError) and e.output:
+                msg = msg + str(e) + "\n"
+                msg = msg + "Subprocess output:"
+                msg = msg + e.output.decode("utf-8", errors="ignore")
+            else:
+                msg = msg + str(e) + ": " + str(tb) + "\n"
         bb.fatal("Fatal errors occurred in subprocesses:\n%s" % msg)
     return results
 

which I think from some local testing gives better output and would solve your concern and some of mine?

Cheers,

Richard

#8 lore raw
Quoted 91 lines · On Thu, Apr 25, 2019 at 3:41 PM <[email protected]> wrote:
Quoted 80 lines · Andrey Zhizhikin
Quoted 64 lines · On Tue, Apr 16, 2019 at 10:24 AM <[email protected]> wrote:
Quoted 57 lines · Andrey Zhizhikin
Quoted 45 lines · Richard Purdie
Quoted 44 lines · Andrey Zhizhikin

Ping.

On Thu, Mar 28, 2019 at 10:47 AM Andrey Zhizhikin < [email protected]

Quoted 40 lines

wrote:
When opkg-build command fails to execute, subprocess is returned
with
exception instead of printing to stderr. This causes the error
logging
not to be printed out, as the "finally" statement does not contain
any
bitbake error output.

One example of this behavior is when the package name contains
uppercase
character, which are rejected by opkg-build, subprocess.check_output
would except and no error log would be produced.

This commit catches the exception
subprocess.CalledProcessError
and
produces bb.error output visible to the user.

---
 meta/classes/package_ipk.bbclass | 2 ++
 1 file changed, 2 insertions(+)
meta/classes/package_ipk.bbclass +0 -0
index d1b317b42b..f181f5b4fd 100644
@@ -234,6 +234,8 @@ def ipk_write_pkg(pkg, d):
             ipk_to_sign = "%s/%s_%s_%s.ipk" % (pkgoutdir,
 pkgname,
 ipkver, d.getVar('PACKAGE_ARCH'))
             sign_ipk(d, ipk_to_sign)
 
Quoted 9 lines
Quoted 5 lines
Quoted 5 lines
meta/classes/package_ipk.bbclass +2 -0
+    except subprocess.CalledProcessError as exc:
+        bb.error("OPKG Build failed: %s" % exc.output)
     finally:

cleanupcontrol(root)
bb.utils.unlockfile(lf)

My main concern is why isn't the raised exception being caught and
causing its own error...

Quoted 5 lines

The raised exception is actually caught by a finally: statement below, and the build gracefully terminates. The problem is that
finally: block does not contain any valuable output to inform user
what actually happened.

Quoted 5 lines

This isn't how python works. The exception should be "re-raised after
the finally clause has been executed" to quote the python manual:
https://docs.python.org/3/tutorial/errors.html#defining-clean-up-actions

Sorry, I guess my previous reply was a bit confusing.. I agree, the exception would not be blocked by finally: statement, and this is why the build gracefully shuts down. What the finally: block does not contain is an bb.error() which would provide more information about the source of error return from subprocess.check_output(). In case if this patch is applied - exception would be handled and not propagated further.

Quoted 8 lines

Can you please advise whether there would another "raise" statement be
needed after bb.error in the patch, so that in addition to the subprocess output user would get an entire callstack (like in the case
when subprocess.CalledProcessError was not handled). Currently, with this patch user would receive the build error with the error string output from subprocess.check_output().

Quoted 8 lines

My worry is that we're making a special case fix and for example the other package back ends could have a similar problem (or any other users of multiprocess_launch).

I already think subprocess in python is a bit broken as it should share e.output if its present in an exception. We already special case that in bitbake, the problem here is that our special case code doesn't catch this.

Agree, I was also puzzled why there is no valuable output from subprocess in case exception was raised.

Quoted 11 lines

Whilst still ugly, perhaps a better fix might be:

meta/lib/oe/utils.py +0 -0
index a4fd79ccb21..59251810d43 100644
@@ -324,7 +324,12 @@ def multiprocess_launch(target, items, d, extraargs=None):
     if errors:
         msg = ""
         for (e, tb) in errors:

meta/lib/oe/utils.py def multiprocess_launch(target, items, d, extraargs=None): · line 335

Quoted 10 lines
meta/lib/oe/utils.py +6 -1
-            msg = msg + str(e) + ": " + str(tb) + "\n"
+            if isinstance(e, subprocess.CalledProcessError) and e.output:
+                msg = msg + str(e) + "\n"
+                msg = msg + "Subprocess output:"
+                msg = msg + e.output.decode("utf-8", errors="ignore")
+            else:
+                msg = msg + str(e) + ": " + str(tb) + "\n"
         bb.fatal("Fatal errors occurred in subprocesses:\n%s" % msg)
     return results
 

Thanks a lot, I'll definitely give it a try! Would you be willing to take this further in into the master branch?

Quoted 3 lines

which I think from some local testing gives better output and would solve your concern and some of mine?

That would definitely solve my issue and adhere to my original intention.

Signature · 2 lines
Regards,
Andrey.
#9 lore raw
Quoted 3 lines · Andrey Zhizhikin

Thanks a lot, I'll definitely give it a try! Would you be willing to take this further in into the master branch?

Just saw your other patch against [utils/multiprocess_launch], please discard this question.

--
Regards,
Andrey.