Skip to contentSkip to conversation

oeqa/selftest/systemd_boot: Never use cleansstate in QA tests without special handling of SSTATE_DIR

PATCH In review Richard Purdie 2 participants 3 messages Jul 15, 2018

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 oeqa/selftest/systemd_boot: Never use cleansstate in QA tests without special handling of SSTATE_DIR

Removing things straight from the live sstate directory is prone to issues since other builds may be relying on the artefact presence.

Also, cleansstate is very slow on the huge sstate that the autobuilder has on slow NFS drives. This may well be causing long buildtimes in oe-selftest as the time taken to remove the artefact by be long.

---
 meta-yocto-bsp/lib/oeqa/selftest/cases/systemd_boot.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
meta-yocto-bsp/lib/oeqa/selftest/cases/systemd_boot.py +1 -1
index dd5eeec1633..26c775c09e9 100644
@@ -44,7 +44,7 @@ class Systemdboot(OESelftestTestCase):
 
         # Ensure we're actually testing that this gets built and not that
         # it was around from an earlier build
-        bitbake('-c cleansstate systemd-boot')
+        bitbake('-c clean systemd-boot')
         runCmd('rm -f %s' % systemdbootfile)
 
         self._common_build()
Signature · 1 line
2.17.1
#1 lore raw
Quoted 6 lines · On 15 July 2018 at 16:17, <[email protected]> wrote:

Removing things straight from the live sstate directory is prone to issues since other builds may be relying on the artefact presence.

Also, cleansstate is very slow on the huge sstate that the autobuilder has on slow NFS drives. This may well be causing long buildtimes in oe-selftest as the time taken to remove the artefact by be long.

Argh, I thought we vetted selftest for this sort of thing already.

Does the test still do the right thing with this? Several instances where it really wanted to build in the past I've replaced with bitbake foo -C unpack or similar.

Ross

#2 lore raw
Quoted 5 lines · Ross
Quoted 4 lines · On 15 July 2018 at 16:17, <[email protected]> wrote:

Removing things straight from the live sstate directory is prone to issues
since other builds may be relying on the artefact presence.

Quoted 7 lines
Quoted 5 lines

Also, cleansstate is very slow on the huge sstate that the autobuilder has
on slow NFS drives. This may well be causing long buildtimes in oe- selftest
as the time taken to remove the artefact by be long.

Argh, I thought we vetted selftest for this sort of thing already.

This one was added in the last 12 months since its in sumo/rocko but not before so we missed it :(.

Quoted 3 lines

Does the test still do the right thing with this? Several instances where it really wanted to build in the past I've replaced with bitbake foo -C unpack or similar.

As far as I can tell it wanted to ensure the item ends up deployed, in which case the -c clean and then the build should be fine. Whether it comes from sstate or not I'm not sure is relevant but I didn't look really closely at the test.

Cheers,

Richard