| Title | Messages from assertions sound like bugs in the MPS |
| Status | closed |
| Priority | essential |
| Assigned user | Richard Brooksby |
| Organization | Ravenbrook |
| Description | When an MPS assertion fires for a client it's almost certainly because the MPS has detected a defect in their code. This is why we do so much checking in the first place. But the MPS messages make it sound like the MPS itself is at fault, saying "MPS ASSERTION FAILED". This is misleading (and bad for us). |
| Analysis | The message could perhaps say "The MPS detected a problem". MPS assertions are a good thing -- finding defects in memory management early in the development process. We should make sure they seem that way, and encourage clients to think positively about them. "The MPS is helping you debug your code." Bruce Mitchener suggested on IRC that the assertion failure could point to the "Common assertions and their causes" section of the manual [2]. Note that if we change the assertion text we have to change the code in [3] that detects this text. |
| How found | customer |
| Evidence | [1] In external integrator recently reported "bad errors" in the MPS, without details <https://info.ravenbrook.com/mail/2013/05/24/17-18-00/0/>.[2] < http://www.ravenbrook.com/project/mps/...html#common-assertions-and-their-causes>[3] < http://info.ravenbrook.com/project/mps/master/test/test/script/headread> |
| Observed in | 1.111.0 |
| Created by | Richard Brooksby |
| Created on | 2013-05-25 22:40:22 |
| Last modified by | Gareth Rees |
| Last modified on | 2014-05-12 15:59:21 |
| History | 2013-05-25 RB Created. |
| Change | Effect | Date | User | Description |
|---|---|---|---|---|
| 186032 | closed | 2014-05-12 15:59:21 | Gareth Rees | Improve the assertion output so that it is less suggestive of a bug in the MPS and more suggestive of a problem that needs investigation. |