This repository has been archived by the owner on Feb 14, 2020. It is now read-only.
ndt.go: send at least one MSG_RESULT back #18
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
The [specification](https://github.com/ndt-project/ndt/wiki/NDTProtocol) doesn't explicitly state that the server MUST send to the client at least one MSG_RESULT message. But, looking into the sequence diagram that is part of the specification, this is CLEARLY implied. As such, botticelli was broken because it was violating the protocol. The proper fix would be to read Web100 variables, compute a number of metrics and send those back to the client. But, botticelli has not reached this point of maturity yet. As a fix, I implemented sending back a single variable formatted as the NDT code would expect, so that it does not crash, but named using a clearly nonstandard name such that it's not used. Bug reported by @nkinkade, thanks!
This was referenced Jan 7, 2017
nkinkade
added a commit
to m-lab/ndt
that referenced
this pull request
May 9, 2017
* Don't assume server supports all test types The NDT client assumes that the server supports all test types and the official NDT server respects this assumption. However, the [neubot/botticelli](https://github.com/neubot/botticelli) server only implements TEST_META, TEST_S2C, and TEST_S2C. Therefore, when using the NDT client with a botticelli server, the client crahes in processing the results of the TEST_MID that however has not been executed. Fix by making sure that, when processing results, we use a bitmak where only bits corresponding to tests that run are actually set. (There is also another reason why NDT client crashes when testing with botticelli, addressed by neubot/botticelli#18.) Problems between NDT client and botticelli reported by @nkinkade, thanks! * web100clt.c: make results parsing more robust 1) Do not assume that the first line we receive contains a space 2) Do not assume that the first line we receive contains an integer 3) Be robust to the case where input is an empty string 4) Do not assume that after the first token delimited by space we will find a second token delimited by newline Tested under the following conditions: a) web100clt -n ndt.iupui.mlab1.trn01.measurement-lab.org that sends back all variables and checked via printf() that the variables that are parsed by the new code are the ones received in resultstr b) web100clt -n neubot.mlab.mlab1.mil01.measurement-lab.org that at the moment is running botticelli v0.0.5 (which is buggy and doesn't send any MSG_RESULTS messages) and make sure it does not crash c) web100clt -n neubot.mlab.mlab1.trn01.measurement-lab.org that at the moment is running botticelli v0.0.6 (which sends a single dummy variable not considered by NDT) and make sure it doesn't crash Note that a) and c) did not changed after this patch. What this patch changes is the behavior in case b). xref: neubot/botticelli#18. * Add travis config to build with measurementlab/ndt-builder * Fetch docker image from mlab-pub * Add log_first_n, cleanup * Add missing #include and make some functions static. * remove confusing I2util discovery * remove web100 dependent test from TESTS * remove unused tcp_stat_setbuf decl that breaks strict lint * Add log_first_n, cleanup * Add missing #include and make some functions static. * Add travis config to build with measurementlab/ndt-builder * Fetch docker image from mlab-pub * remove confusing I2util discovery * remove web100 dependent test from TESTS * remove unused tcp_stat_setbuf decl that breaks strict lint * Temp workaround for builder * Remove logfile after test (#88) * Made the unit tests create the logfile in /tmp * Clarified the test code
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
The specification doesn't explicitly state that the server MUST send to the client at least one MSG_RESULT message. But, looking into the sequence diagram that is part of the specification, this is CLEARLY implied.
As such, botticelli was broken because it was violating the protocol.
The proper fix would be to read Web100 variables, compute a number of metrics and send those back to the client. But, botticelli has not reached this point of maturity yet.
As a fix, I implemented sending back a single variable formatted as the NDT code would expect, so that it does not crash, but named using a clearly nonstandard name such that it's not used.
Bug reported by @nkinkade, thanks!