[LU-5012] newwiretest is not run automatically by build or test Created: 06/May/14 Updated: 13/May/14 |
|
| Status: | Open |
| Project: | Lustre |
| Component/s: | None |
| Affects Version/s: | Lustre 2.6.0 |
| Fix Version/s: | None |
| Type: | Bug | Priority: | Minor |
| Reporter: | John Hammond | Assignee: | WC Triage |
| Resolution: | Unresolved | Votes: | 0 |
| Labels: | build, test | ||
| Severity: | 3 |
| Rank (Obsolete): | 13873 |
| Description |
|
newwiretest is not run automatically by build or test. |
| Comments |
| Comment by Andreas Dilger [ 06/May/14 ] |
|
This is not a bug. The whole point of wiretest is that this test should not change unless some user is deliberately modifying the wire protocol. Otherwise, if this file is regenerated automatically the wire protocol may change without sufficient notification. |
| Comment by John Hammond [ 06/May/14 ] |
|
So we're on the honor system to use it? And to use it properly? |
| Comment by John Hammond [ 06/May/14 ] |
|
Let me clarity what I meant in the description. Assuming that newwiretest was modified (or cloned) so that it didn't always (or didn't) clobber source files, then shouldn't that non modifying version always pass? |
| Comment by Andreas Dilger [ 13/May/14 ] |
|
I'm still not sure I understand the question. It doesn't make sense to me that a newly-generated wiretest could fail, since it will always be checking "FOO == FOO". The purpose of wiretest is to make sure that the value of "FOO" does not change over time. Note that there is a sanity.sh test that runs wiretest that is run on any system test, along with one in ptlrpc that is run on every mount. Since all of these checks should optimize away at compile time it doesn't add any overhead. These checks do not catch many problems today, but in the past they caught many problems. |