[
https://issues.apache.org/jira/browse/GEODE-8756?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17251972#comment-17251972
]
Mario Salazar de Torres commented on GEODE-8756:
------------------------------------------------
Hi [~bbender],
Sorry for any trouble. Sadly I do not have available right now a RHEL7
environment, but I wll get my hands on one.
I wanted to point out that thanks to [~moleske] we've improved a lot the
testing for PRs, still I don't know if it's possible for you to attach a list
of the platforms you run the tests in so we can have a similar setup. Things
that would be nice to know for each testing environment:
- OS
- Compiler.
- Test executed.
Will open a new PR once the RHEL7 issue is sorted.
BR,
Mario.
> CacheableString objectSize is not correct
> -----------------------------------------
>
> Key: GEODE-8756
> URL: https://issues.apache.org/jira/browse/GEODE-8756
> Project: Geode
> Issue Type: Sub-task
> Components: native client
> Affects Versions: 1.11.0, 1.12.0, 1.13.0, 1.13.1
> Reporter: Mario Salazar de Torres
> Assignee: Mario Salazar de Torres
> Priority: Major
> Labels: pull-request-available
>
> CacheableString objectSize function is returning an incorrect value.
> This class is based upon STL's string implementation, and most of the
> compilers implementations apply what's called SSO.
> What SSO basically does is if the string occupies less than a certain amount,
> no extra memory would be allocated in the heap, and the character-sequence
> would be stored in the object itself. This is typically achieved by using
> union semantics.
> Right now if SSO applies, objectSize calculates the size of std::string as
> sizeof(std::string) + m_str.capacity(), which is more than it actually
> occupies.
> On the other hand starting C++11 STL's strings needs to allocate an extra
> character
> to keep the null-terminator in the same buffer as the actual string. This is
> specified in section § 21.4.7.1 within the C++11 standard.
> Because of this objectSize should take the null-terminator into account,
> which was not the case.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
