Attempt to fix a probable bug introduced by GPR#156.#279
Attempt to fix a probable bug introduced by GPR#156.#279alainfrisch merged 1 commit intoocaml:trunkfrom
Conversation
Attempt to fix a probable bug introduced by GPR#156.
|
We've got confirmation that this patch fixes the segfaults on several tests for our customer, so I've merged this PR.
@mshinwell @chambart Can you confirm that the patch proposed here (now merged) fixes those segfaults? |
|
@alainfrisch The fix seems correct. In fact it was missing only missing There is no need to move the last expressions into the if branches (frametables is correctly NULL at that point in the first branch) Sorry for the bug, it seems that I introduced it while renaming things in the patch chambart@3fdaf02#diff-3c7717e893a739a54e5fbc92ec4b7100L148 |
I don't understand. Since frametables is NULL, |
|
Effectively, sorry... |
|
Is this solved to everyone's satisfaction? If so please close this PR as well as PR#156. |
|
I closed #156, but for some reason, I don't see a "Close" button on this one. |
|
It is automatically considered closed (in fact "Merged") since you merged the corresponding patch -- this PR does not appear in the list of open PRs anymore. |
|
Ah thanks, I was confused by Xavier's request. |
|
My bad, I was probably looking at an outdated Web page generated by Github. |
Fix assert in caml_ephe_clean
* Various improvements related to unboxing and meet * Review: Catch invalid field reads earlier * Add comment on Row_like.For_blocks.get_field
See my last note on #156.
It would be interesting to re-run benchmarks done when accepting #156 to see if the gains are confirmed after this fix.