RFR: JDK-8260902: CDS mapping errors should not lead to unconditional output

[Ioi Lam]

On Tue, 2 Feb 2021 09:16:47 GMT, Thomas Stuefe <stuefe at openjdk.org> wrote:

> There are some places where unconditional error messages are printed to tty for recoverable errors. This is not good since the VM may continue without problems, but those printouts may confuse scripts parsing the VM output, or end up at the wrong places:
> 
> See e.g.: https://mail.openjdk.java.net/pipermail/build-dev/2021-February/030256.html
> 
> One of these points are mapping errors in CDS. VM will just continue without CDS, so its not a fatal error. 
> 
> I fixed some places which clearly were non-fatal while leaving log_error in place for cases which are clearly fatal, or borderline fatal (like rs commit errors). I looked for tests which could have parsed these lines and found none.

Changes requested by iklam (Reviewer).

src/hotspot/share/memory/filemap.cpp line 1675:

> 1673: 
> 1674:   if (VerifySharedSpaces && !region_crc_check(bitmap_base, si->used_aligned(), si->crc())) {
> 1675:     log_info(cds)("relocation bitmap CRC error");

I think this one should remain an error -- it means the CDS archive has been corrupted.

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

PR: https://git.openjdk.java.net/jdk/pull/2348