RFR: 8204680: Disassembly does not display code strings in stubs
stuart.monteith at linaro.org
Fri Jun 15 18:02:16 UTC 2018
There appears to be a bug introduced here. I've opened
https://bugs.openjdk.java.net/browse/JDK-8205118 and I am
On 11 June 2018 at 21:35, Vladimir Kozlov <vladimir.kozlov at oracle.com> wrote:
> Looks fine to me.
> On 6/11/18 7:37 AM, Andrew Haley wrote:
>> So last Friday I was looking at the code we generate for the runtime
>> stubs and I noticed that there were no comments in the disassembly.
>> Which is odd, because I'm sure it used to work. I found a bug which
>> prevented it from working, fixed it, but there was still no output.
>> What??! This led me down a rabbit hole from which I was to emerge
>> several hours later.
>> It turns out there are two separate bugs.
>> When we disassemble, the code strings are found in the CodeBlob that
>> contains the code. Unfortunately, when we use -XX:+PrintStubCode the
>> disassembly is done from a CodeBuffer before the code strings have
>> actually been copied to the code blob, so the disassembler finds no
>> code strings.
>> Also, the code strings are only copied into the CodeBlob if
>> PrintStubCode is true, so "call disnm()" in the debugger doesn't print
>> any code strings because they were lost when the CodeBlob was created.
>> With both of these fixed, we have fully-commented disassembly in the
>> stubs again.
More information about the hotspot-dev