On Wed, 20 May 2026 10:59:50 GMT, Jaikiran Pai <[email protected]> wrote:
>> SendaoYan has updated the pull request with a new target base due to a merge
>> or a rebase. The incremental webrev excludes the unrelated changes brought
>> in by the merge/rebase. The pull request contains two additional commits
>> since the last revision:
>>
>> - Merge branch 'openjdk:master' into jbs8381851
>> - 8381851: handleMessage use uninitialized struct
>
> src/jdk.sctp/unix/native/libsctp/SctpChannelImpl.c line 392:
>
>> 390: jboolean isEOR, struct sockaddr* sap) {
>> 391: jobject isa, resultObj;
>> 392: struct controlData cdata[1] = {0};
>
> Initializing each of the fields on the `controlData` struct instance to `0`,
> looks OK to me.
>
> A larger question however is how this code should deal with the case where
> the implementation in `getControlData(msg, cdata)` doesn't set/overwrite this
> instance's field values i.e. if there are no messages which have
> `cmsg->cmsg_level == IPPROTO_SCTP && cmsg->cmsg_type == SCTP_SNDRCV`. It
> looks like we don't check for that possibility after we return from that call
> to `getControlData()` and go ahead and construct the MessageInfoImpl with the
> now default values of `0` for these fields. But that part of the code is
> pre-existing and your proposed change here in fact improves the situation by
> setting it to a default value.
I am not familiar to those sctp codes. Maybe there are always have at least one
message which have `cmsg->cmsg_level == IPPROTO_SCTP && cmsg->cmsg_type ==
SCTP_SNDRCV`, so there is no sctp bug reported for now. Or maybe little people
use sctp through java?
Anyway, initial the struct variable do avoid jvm crash which build by
clang23/llvm23, and the sctp tests run passed.
-------------
PR Review Comment: https://git.openjdk.org/jdk/pull/30664#discussion_r3274002418
