bitbake: main: Alter EOFError handling

If the server shuts down 'cleanly' due to some issue, the socket will close.
A recently reported example was an invalid PRSERV being set. Doing this silently
and without changing the retries count will case the server startup to loop
infinitely.

Change the code so it triggers the usual retries note messages and times
out eventually pointing the user at the cooker log file.

[YOCTO #12984]

(Bitbake rev: bb696636ef0c59f9e9640bb9460e7cce323cc785)

Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This commit is contained in:
Richard Purdie
2019-07-02 11:56:31 +01:00
parent 53138165bc
commit 97e19bf73b

View File

@@ -447,12 +447,7 @@ def setup_bitbake(configParams, configuration, extrafeatures=None):
bb.utils.unlockfile(lock)
raise bb.server.process.ProcessTimeout("Bitbake still shutting down as socket exists but no lock?")
if not configParams.server_only:
try:
server_connection = bb.server.process.connectProcessServer(sockname, featureset)
except EOFError:
# The server may have been shutting down but not closed the socket yet. If that happened,
# ignore it.
pass
server_connection = bb.server.process.connectProcessServer(sockname, featureset)
if server_connection or configParams.server_only:
break
@@ -463,7 +458,7 @@ def setup_bitbake(configParams, configuration, extrafeatures=None):
raise
retries -= 1
tryno = 8 - retries
if isinstance(e, (bb.server.process.ProcessTimeout, BrokenPipeError)):
if isinstance(e, (bb.server.process.ProcessTimeout, BrokenPipeError, EOFError)):
logger.info("Retrying server connection (#%d)..." % tryno)
else:
logger.info("Retrying server connection (#%d)... (%s)" % (tryno, traceback.format_exc()))