On Wed, 03/21 08:14, Eric Blake wrote: > On 03/21/2018 03:11 AM, Fam Zheng wrote: > > > Thanks for this; today was my first day trying the various vm-build- > > > targets. > > > > > > Question: is this expected behavior? > > > > > > $ make vm-build-ubuntu.i386 > > > VM-IMAGE ubuntu.i386 > > > ... > > > Image resized. > > > debconf: unable to initialize frontend: Dialog > > > debconf: (TERM is not set, so the dialog frontend is not usable.) > > > > > > > I'm wondering if the image initialized incorrectly, and as a result can't > > > do > > > as much as it's supposed to do, or runs way slower than it needs to? The > > > command eventually completed with status 0, but failed to find a 32-bit > > > compile error in the rdma code, and the output log does not look like it > > > actually attempted to compile anything in qemu (unless it did compile it, > > > but the logs were not output to stdout/stderr). > > > > > > make vm-build-freebsd was a lot faster at completing for me (but shows > > > that > > > we still have a lot of clang warnings about address of a packed struct > > > member). > > > > > > > Was it a clean repo? Did you try "rm -rf ~/.cache/qemu-vm"? This morning the > > command worked for me from a clean env (on RHEL 7, using a QEMU built from > > qemu.git), and I didn't see the errors/warnings you pasted. (BTW I don't > > understand how sudo has anything to do with the dummy host name.) > > I repeated the test twice this morning after 'rm -rf ~/.cache/qemu-vm', > first after 'rm tests/vm/ubuntu.i386.img' in an incremental tree, and second > in a fresh git checkout. Here's a full pastebin of the first try (I hit ^C > after seeing it look the same as yesterday, even after the fresh redownload > that took more than 10 minutes): > https://paste.fedoraproject.org/paste/p8Th7AroLAprGY03ffzGHA > > The results are the same; debconf is somehow unable to connect to the guest, > and as a result, the guest is not properly initialized with the additional > packages it needs, and then everything else about the image is broken. > > Then on the second try, I got a dreaded message about my /home being nearly > full, (having multiple .img files in two different build directories pushed > me over limits), and it died a much harder death due to ENOSPC. Let me > resize my /home and retry (thankfully I still have some disk space free to > throw at the problem), and I'll report back if that makes a difference.
Now I see the problem. I always use V=1 so stdout is different for the spawned commands (qemu and ssh) in the script. I will send a patch tomorrow. Thanks! Fam
