[Starlingx-discuss] Build Packages: Fail to Build BASH with Simultaneous Jobs
During a recent compilation of StarlingX build failures happened for package bash-4.2.46-29 [0] 2 ways to fix: - in a workstation with 8 CPUs, it is fixed by cleaning the component and running a second time via build-pkgs - in a server with 88 CPUs, the failure is fixed by removing the -j option from patch [1] Any advice is highly appreciated. Abraham [0] https://github.com/starlingx-staging/stx-gplv2/tree/master/bash [1] https://github.com/starlingx-staging/stx-gplv2/blob/master/bash/centos/meta_...
Neither seems good. It sounds like there is a race condition in the build. Can we fix that? Has this been fixed in newer versions of bash? brucej From: Arce Moreno, Abraham [mailto:abraham.arce.moreno@intel.com] Sent: Wednesday, June 13, 2018 2:47 PM To: starlingx-discuss@lists.starlingx.io Subject: [Starlingx-discuss] Build Packages: Fail to Build BASH with Simultaneous Jobs During a recent compilation of StarlingX build failures happened for package bash-4.2.46-29 [0] 2 ways to fix: - in a workstation with 8 CPUs, it is fixed by cleaning the component and running a second time via build-pkgs - in a server with 88 CPUs, the failure is fixed by removing the -j option from patch [1] Any advice is highly appreciated. Abraham [0] https://github.com/starlingx-staging/stx-gplv2/tree/master/bash [1] https://github.com/starlingx-staging/stx-gplv2/blob/master/bash/centos/meta_...
Neither seems good. It sounds like there is a race condition in the build. Can we fix that? Has this been fixed in newer versions of bash?
Thanks Bruce! Let me annotate the following options and I will complement with Chris' options: 1. Fix the problem 2. Update Bash package Abraham
On 06/13/2018 03:46 PM, Arce Moreno, Abraham wrote:
During a recent compilation of StarlingX build failures happened for package bash-4.2.46-29 [0] 2 ways to fix: - in a workstation with 8 CPUs, it is fixed by cleaning the component and running a second time via build-pkgs - in a server with 88 CPUs, the failure is fixed by removing the -j option from patch [1] Any advice is highly appreciated.
The patch in question added -j"%(nprocs)" to the "make" line as part of an attempt to speed up the overall build on machines with many CPUs. The fact that it's failing points to a flaw in the upstream package. I see two options, either track down and fix the problem with parallel builds or else remove the "-j" option until the upstream package fixes parallel builds. Chris
During a recent compilation of StarlingX build failures happened for package bash-4.2.46-29 [0] 2 ways to fix: - in a workstation with 8 CPUs, it is fixed by cleaning the component and running a second time via build-pkgs - in a server with 88 CPUs, the failure is fixed by removing the -j option from patch [1] Any advice is highly appreciated.
The patch in question added -j"%(nprocs)" to the "make" line as part of an attempt to speed up the overall build on machines with many CPUs. The fact that it's failing points to a flaw in the upstream package.
Yes, we have had no other package has failed due to this option.
I see two options, either track down and fix the problem with parallel builds or else remove the "-j" option until the upstream package fixes parallel builds.
Now complementing with Bruce's options: 1. Update Bash package 2. Fix the problem 3. Remove patch I will work as they are listed, assuming no other package is tightly coupled to bash version 4.2.46. Would this upgrade be a good first exercise to allow StarlingX have latest package versions? Abraham
participants (3)
-
Arce Moreno, Abraham
-
Chris Friesen
-
Jones, Bruce E