[FAB-18509] Stop panic if collection index path is wrong #2726
Conversation
manish-sethi
changed the title
| @@ -420,21 +420,26 @@ const ( | |||
| // Example for chaincode indexes: | |||
| // "META-INF/statedb/couchdb/indexes/indexColorSortName.json" | |||
| chaincodeIndexDirDepth = 3 | |||
| pathLengthForCCindexes = 5 | |||
There was a problem hiding this comment.
| pathLengthForCCindexes = 5 | |
| pathLengthForCCindexes = 4 |
Path that comes here is of directory only. The file name is not part of this.
There was a problem hiding this comment.
thanks for the review Manish - this is the output I got from the unit test - with printing the args to the getIndexInfo function. Includes the index file name
=== RUN TestGetIndexInfo
[getIndexInfo] indexPath=META-INF/statedb/couchdb/indexes/indexColorSortName.json
[getIndexInfo] indexPath=META-INF/statedb/couchdb/collections/collectionMarbles/indexes/indexCollMarbles.json
[getIndexInfo] indexPath=META-INF/statedb/couchdb/indexColorSortName.json
[getIndexInfo] indexPath=META-INF/statedb/couchdb/collections/indexes/indexCollMarbles.json
[getIndexInfo] indexPath=META-INF/statedb/couchdb/collections/indexCollMarbles.json
[getIndexInfo] indexPath=META-INF/statedb/
--- PASS: TestGetIndexInfo (0.00s)
There was a problem hiding this comment.
@mbwhite Existing TestGetIndexInfo() has a bug too. It should not pass the path including the index def, i.e., *.json. The input to getIndexInfo() comes from the HandleDeployChaincode() which passes only the directory excluding the actual filename.
We need to fix
TestGetIndexInfo()- Comments (provided in the const block)
getIndexInfo()
There was a problem hiding this comment.
Some tests in kvledger package and integration test also fail after making getIndexInfo() to consider the full path including the index file.
There was a problem hiding this comment.
@cendhu ah.. that makes more sense now thanks... what can we say about the directory path that is passed in?
The code as-is is uses '/' to split the path, is it safe to assume the input will not have leading or trailing directory separators?
(aside I assume as well that Fabric is solely intended to work on a Linux filesystem? / not being the universal file separator :-) )
There was a problem hiding this comment.
ExtractFileEntries(). I do not see a unit test for ExtractFileEntries(). Not sure whether it is tested somewhere and I miss it. Hence, it is better to verify that.
As far as I know, even on Windows, we recommend docker to be used (https://hyperledger-fabric.readthedocs.io/en/release-1.4/prereqs.html) and our docker base image is alpine linux. Hence, / as a file separator is okay.
There was a problem hiding this comment.
After a bit more research I believe there are some in db_test.go. My reading of that function is that it later functions can safely assume there are no leading or trailing /
My background in cross-platform development (pre Docker) has made me question whenever I see anything about a filesystem 'hard coded' like a /
There was a problem hiding this comment.
I see. Thanks for confirming.
| collectionDirDepth = 3 | ||
| collectionNameDepth = 4 | ||
| collectionIndexDirDepth = 5 | ||
| pathLengthForCollectionindexes = 7 |
There was a problem hiding this comment.
| pathLengthForCollectionindexes = 7 | |
| pathLengthForCollectionindexes = 6 |
same as above
mbwhite
force-pushed
the
index-dir-checks
branch
3 times, most recently
from
9866539 to
d5fedc0
Compare
|
/ci-run |
Signed-off-by: Matthew B White <whitemat@uk.ibm.com>
|
@Mergifyio backport release-2.2 |
|
@Mergifyio backport release-2.3 |
|
Command
|
Signed-off-by: Matthew B White <whitemat@uk.ibm.com> (cherry picked from commit 1249da2)
|
Command
|
Signed-off-by: Matthew B White <whitemat@uk.ibm.com> (cherry picked from commit 1249da2)
…#2726) (hyperledger#2745) Signed-off-by: Matthew B White <whitemat@uk.ibm.com> (cherry picked from commit 1249da2)
Signed-off-by: Matthew B White whitemat@uk.ibm.com
Type of change