Hi Suenaga-san,
I think the return value from the second MultiByteToWideChar invocation
should be examined (non-zero), and if it was not successful, appropriate
action should be taken.
Naoto
On 3/9/20 3:50 AM, Yasumasa Suenaga wrote:
Hi all,
Please review this change:
JBS: https://bugs.openjdk.java.net/browse/JDK-8240725
webrev: http://cr.openjdk.java.net/~ysuenaga/JDK-8240725/webrev.00/
We found the issue that HotSpot does not start when it is deployed on
the path which contains CJK character(s), and it has been fixed in
JDK-8240197.
On the review of JDK-8240197 [1], we concern similar issue might occur
in other place, and I found potentially problem in below:
- ZFILE_Open() @ zip_util.c
- JDK_Canonicalize() @ canonicalize_md.c (for Windows)
- create_unc_path() @ java_md.c (for Windows)
- Platform::MultibyteStringToWideString() @ WindowsPlatform.cpp
This change passed tests on submit repo
(mach5-one-ysuenaga-JDK-8240725-20200309-0811-9304139).
Thanks,
Yasumasa
[1]
https://mail.openjdk.java.net/pipermail/hotspot-runtime-dev/2020-March/038397.html
