Integration Test #79
Integration Test #79
Conversation
g2flyer
force-pushed
the
msteiner.integration-test
branch
from
a057d00
to
de1685d
Compare
There was a problem hiding this comment.
I get an error. see here
auction_test.sh: FIRST RUN: this should fail due to docker-switcheroo ...
auction_test.sh: - setup ledger
/home/bur/fabric-secure-chaincode/integration/common_ledger.sh: line 56: kill: (4565) - No such process
auction_test.sh: Orderer quit too quickly: (for log see /tmp/hyperledger/test//orderer.out & /tmp/hyperledger/test//orderer.err)
/home/bur/fabric-secure-chaincode/integration/common_ledger.sh: line 96: kill: (4565) - No such process
Makefile:14: recipe for target 'auction_test' failed
de1685d
to
b439b41
Compare
I saw the same problem yesterday also. I'm not sure what exactly made it fail but making fabric config and cc-name etc test-specific seemed to remove the problem. Test again with the new version. BTW: this should hopefully be obsolete anyway soon: i'm mid-way through a patch which removes the switcheroo (or more proactively, creates the correct container before the peer generates the wrong one ...) |
|
I am also running into that problem. Any updates? |
You were running with the latest version? I couldn't reproduce the issue with that version. Actually, thinking of it i'm actually not sure i ever saw exactly that problem. What do the log-files show (see the error message for location of log-files)? |
|
alright, I was missing credentials in filed an issue #82 for this: if you want to address it now fine, if you want to address it later good! |
g2flyer
force-pushed
the
msteiner.integration-test
branch
from
b439b41
to
e58d4a3
Compare
Updated the PR with corresponding tests and a bit of generalizations (see #82 for comments on how the common_* scripts can (and should) be moved to a global scope as part of the SDK refactoring ..) |
There was a problem hiding this comment.
I get few errors when I run make.
First - linter complains.
fabric-private-chaincode git:(g2flyer-msteiner.integration-test) ✗ pwd
/home/bur/go/src/github.com/hyperledger-labs/fabric-private-chaincode
➜ fabric-private-chaincode git:(g2flyer-msteiner.integration-test) ✗ make checks
License: Running licence checks..
head: error reading 'fabric/sgxconfig/ias': Is a directory
head: error reading 'integration/config/ias': Is a directory
grep: fabric/sgxconfig/ias: Is a directory
grep: integration/config/ias: Is a directory
The following files are missing SPDX-License-Identifier headers:
config/ias/.gitignore
fabric/sgxconfig/ias
.gitignore
integration/config/ias
Please replace the Apache license header comment text with:
SPDX-License-Identifier: Apache-2.0
Checking committed files for traditional Apache License headers ...
grep: fabric/sgxconfig/ias: Is a directory
grep: integration/config/ias: Is a directory
The following files are missing traditional Apache 2.0 headers:
config/ias/.gitignore
fabric/sgxconfig/ias
.gitignore
integration/config/ias
Fatal Error - All files must have a license header
Makefile:16: recipe for target 'license' failed
make: *** [license] Error 1
And second, somehow the auction script does not run with my setup. Note that in my $GOPATH/src/github.com/hyperledger-labs/ I have a symlink to /home/bur/fabric-private-chaincode, seems this causes the problem?!?!
➜ integration git:(g2flyer-msteiner.integration-test) ✗ pwd
/home/bur/go/src/github.com/hyperledger-labs/fabric-private-chaincode/integration
➜ integration git:(g2flyer-msteiner.integration-test) ✗ make
./auction_test.sh
auction_test.sh: FABRIC_BIN_DIR not properly defined as '/home/bur/fabric-secure-chaincode/integration/../../../hyperledger/fabric/.build/bin'
Makefile:14: recipe for target 'auction_test' failed
make: *** [auction_test] Error 111
|
Regarding symlinks, in general you should haveonly symlinks into GOPATH, but not out of GOPATH as you seem to do: In this particular case you could define FABRIC_BIN_DIR env variable but i think other things won't work reliabilty if you do not have a proper (bidirectional) GOPATH tree. Just move your directory into GOPATH and then symlink from ~/ and you still have quick access? Regarding license checks failing, clearly these files cannot have a license header. Also, doesn't fail for me. (Anyway, this is a fabric script so you might be anyway the better person to debug ;-) |
integration/auction_test.sh
Outdated
| . ${SCRIPTDIR}/common_utils.sh | ||
| . ${SCRIPTDIR}/common_ledger.sh | ||
|
|
||
| CC_ID=ecc_auction_test |
There was a problem hiding this comment.
It is failing for me and this should be the problem:
you have updated the id but not vscc;
there is also an mrenclave not found -- I am still debugging
There was a problem hiding this comment.
you have updated the id but not vscc;
i only changed id to make docker images distinct if you run multiple setups, e.g., integration test and example, on same machine. In any case, not sure what id you mean when you refer to vscc? This is a reference to code (and not an instance of the chaincode)?
there is also an mrenclave not found -- I am still debugging
not sure what you are referring to here
Actually, now it is failing also but only after i noticed somebody merged master into this branch and I (force) pulled the new version. PS: I'd prefer if unless handing over control i would own my own PR branches and the merge of master and alike (e.g., i would rather have done a rebase than a merge and it is also easier to understand what happens if you report problems) |
|
got it, and fine with that -- just wanted to speed up a possible approval, since it was good before. |
Hmm, didn't you approve before my test for the config (which i took as a confirmation that it works) ? Are you saying the version after todays force-push but before the merge failed but did work before my merge-push with config test? If so, how did it fail? |
|
Yes, it worked before. |
Looking now at the license check, i notice that it actually only tests files (and directories) modified in the current commit and the attempt to filter out files which should not have a license is very brittle. I don't think that this is something we realistically can make robust (and i'm pretty sure it will fail for fabric also on quite some commits), so i would suggest to leave it as is and just for such failing license checks manually peruse the files which make it fail and if bening ignore the errors ... We might, though, remove the dependency on 'check' for target 'test' and move 'check' to last in list for target 'all' in build.mk? Any reason test should depend on checks? |
Nope, nothing changed in terms of id in todays diff. (if you click on the force-push in the history, you get the actual delta pushed, e.g., https://github.com/hyperledger-labs/fabric-private-chaincode/compare/b439b413cb06806b1ccb186aed9656976a6880e8..e58d4a3f515b29990878775cf764b81b70c13285 in this case) |
btw: did you do a 'make clean' before you run the test? |
Of course I did. |
That change is from a week ago (in this force-push/rebase/delta-commit: https://github.com/hyperledger-labs/fabric-private-chaincode/compare/de1685d52ca581aede1c1d76f95ca1cf568faa36..b439b413cb06806b1ccb186aed9656976a6880e8). I've done rebase to add the patches to keep a clean commit history, so just looking at the final commits doesn't tell you the full story .... Anyway, what exactly is the problem/error you get with vscc? |
|
2019-06-13 07:46:02.936 PDT [tl-enclave] golog -> INFO 05a ERROR [Enclave] Ledger: Only SGX chaincodes supported currently! then it crashes. About the changes and commits. I have no idea of what you are talking about. |
Did you click on the links i provided? They correspond (as mentioned) to the incremental changes in introduced in the rebase of the commits you mention. They show up as force-push (links) in the history on this page. |
This sounds like an error in the extraction of the MRENCLAVE. As i pointed out in an earlier PR #75, we didn't do proper return code checking and anyway extract the MRENCLAVE much more complex/brittle than necessary. The latter part is issue #77 the former i thought we fixed but maybe not ... (doesn't seem so, at least *_enclave/generate_mrenclave.sh still doesn't check any errors ..) |
* integration-test dir structure with externalized sgx-config (also for demo) * better logging defaults for peer * a few more common ignores Signed-off-by: Michael Steiner <michael.steiner@intel.com>
g2flyer
force-pushed
the
msteiner.integration-test
branch
from
ddef018
to
f0193b1
Compare
Did add now fix with missing error check (as well as pushing 'checks' make target to last, see above). Hopefully, that helps root-causing problem Bruno faces ... |
* added various missing license headers * integration test based on auction CC Signed-off-by: Michael Steiner <michael.steiner@intel.com>
, though still remaining issue hyperledger#77 on simplifying logic) and also removed duplication of script itself. * make checks target last and test not depending on it as license check is rather brittle and requires often manual error-checking (and ignoring thereof) * small tweak in make checks target to treat both scripts identical Signed-off-by: Michael Steiner <michael.steiner@intel.com>
g2flyer
force-pushed
the
msteiner.integration-test
branch
from
a8ac9ee
to
7dc6d35
Compare
|
It works for me on 7dc6d35 |
|
Great job! Thanks for this PR! |
subject says it all (or mostly: besides address issue #67, i also did some related cleanup / re-org of fabric/sgxconfig/* and addressed issue #53 )
PS: eventually, some of the .sh libraries could also be put into common but for now i left it self-contained ./integration ...