677350 - Move main phase linux/linux-64 build slaves to new linux machine; upgrade OS on old slave host then move linux/linux-64 new sandbox slaves to that machine.

Status: VERIFIED FIXED

(Tamarin Graveyard :: Build Config, defect, P2)

Description

[James Sudduth]

The new sandbox linux/linux-64 slaves have been failing with gcc errors (see bugs 648249, 664143 and 656008 for background on the gcc problem). Those machines are running Red Hat linux and we are not able to upgrade gcc to the right version (4.3 or 4.4) to resolve the errors. 

The fix for now is to move the Main phase linux/linux-64 slaves to our new linux machine, an Ubuntu box on which we can obtain gcc 4.3/4.4. Then we will upgrade Ubuntu and buildbot on the old machine and move the new sandbox linux/linux-64 slaves to that machine.

Comment 1

[James Sudduth]

Attachment: Add linux-arm slave to new sandbox, fix linux-arm-deep environment.sh

Adds a linux-arm slave to the new Buildbot 0.8.3 sandbox. Also fixes a problem in the linux-arm-deep slave where the asteambeagle3 and asteambeagle4 workers were not using both /app1 and /app2 and were using only two threads instead of four.

Comment 2

[James Sudduth]

A change of plan: it turns out we can build gcc 4.3 on the Red Hat boxes (and do some other upgrades) which will allow us to continue using them as slave machines. One has been upgraded and tested, so now the other two will be similarly treated.

Comment 3

[James Sudduth]

Still in progress - the upgrades on the two additional vm's are nearly completed and will be tested today.

Comment 4

[James Sudduth]

After observing the builds for several days to be sure no unexpected problems arise the linux/linux64 builds seem to be fine now on all three VM's.