ffldb: Add dbErr to error description.

[JoeGruffins]

ffldb: Add dbErr to error description.

Add the reason for bad inserts, updates, and deletes to the error
description and display key as hex in order to give the user more
information when the database errors. This is especially useful for
upgrades which may fail for several reasons, such as low disc space,
which the user will now be informed about.

blockchain: Add ldbErr to error description.

Add the reason for bad inserts, updates, and deletes to the error
description.

I am sorry for not making an issue first. I believe displaying the key as hex was lightly discussed in matrix and the consensus was that it's ok.

Error for low disc space on master:

[ERR] DCRD: Unable to start server: failed to put key "\x00\x00\x00\x19\x1dn\xf1\x14()d\x17*\xa2\x91y\x9c\xcc\xf5Ic{(\xbaW\xbc\xe3\x92G\x00\x00\x00\x00\x00\x00\x00" to ldb transaction

With this diff:

[ERR] DCRD: Unable to start server: failed to put key 0000001620a9ba256242f74a82afe270bacaae88680e636b698cca830500000000000000 to ldb transaction: write /home/joe/.dcrd/data/mainnet/blocks_ffldb/metadata/069617.ldb: no space left on device

Tested with a real upgrade.

[rstaudt2]

Looks like a good improvement to me. The underlying error is being stored in the RawErr field of the converted error, but that field is never displayed in the error output.

Regarding displaying the key as hex, we did have a brief discussion of it in this PR for a similar error case.

If you are updating this location, it probably makes sense to update the case below for deletions as well. Or, another option could be to update the convertErr function to include the error to handle it for all cases.

[JoeGruffins]

Now returning more info for all errors. Unsure how to trigger others besides low disc space, but that one looks the same.

[rstaudt2]

Now returning more info for all errors. Unsure how to trigger others besides low disc space, but that one looks the same.

The update looks good to me. Just a couple of minor comments:

• It would be good to make the same update at this location for the UTXO backend as well (could be the same or separate commit)
• It is probably worth updating the function comment to indicate that the original error is included in the description so that the caller is aware

[JoeGruffins]

doh! I shoulda run tests.

I realized this caused some of the errors to print twice, as it was using the error's Error for the description. Gave those different descriptions that say a little more about where the error took place. Will fix tests.

[rstaudt2]

Tested both locations by updating the leveldb code to force an error to be returned, and it worked as expected.

[davecgh]

The code changes look good and I also tested various failures by intentionally inducing them. Showing the underlying error is nice as it can often make it more obvious why something failed in cases such as out of disk space errors.