RFR: 8328630: Add logging when needed symbols in dll are missing. [v2]

David Holmes dholmes at openjdk.org
Mon Mar 25 01:18:22 UTC 2024


On Fri, 22 Mar 2024 10:23:37 GMT, Robbin Ehn <rehn at openjdk.org> wrote:

>> The result are stored in per thread struct, so that method just load that thread local and check that result is NULL and returns. IMHO that's fine.
>> 
>> If there is an error it must free it, so maybe considering all the libjsvml ~40-50 errors that must be freed ?
>> 
>> Thinking about it, those 40-50 failed lookups are much, much, more expensive.
>> So it would make more sense to fix that side.
>
> Here is the data: Avg:187.913 ns Sum:40965 ns
> 
> Want me to add the guard? I don't mind.

Sorry Robbin can you just re-explain why we need the initial call to `dl_error` here please. If the `dl_sym` is successful then we shouldn't be checking `dl_error`; and if not successful then `dl_error` should be returning that most recent failure.

-------------

PR Review Comment: https://git.openjdk.org/jdk/pull/18404#discussion_r1536963440


More information about the hotspot-runtime-dev mailing list