RFR(S) 8204294: [REDO] - JVMFlag::printError missing ATTRIBUTE_PRINTF

[Kim Barrett]

> On Sep 14, 2018, at 1:40 PM, Gerard Ziemski <gerard.ziemski at oracle.com> wrote:
> 
> Hi all,
> 
> Please review this redo of a simple fix that adds ATTRIBUTE_PRINTF to "JVMFlag::printError” and also fixes the resulting mismatch format errors that it triggers (as it is supposed to).
> 
> Bug: https://bugs.openjdk.java.net/browse/JDK-8204294
> Webrev: http://cr.openjdk.java.net/~gziemski/8204294_rev1/index.html
> Testing: Mach5 hs_tier1,2,3,4,5,6 (completed successfully)
> 
> 
> Cheers

------------------------------------------------------------------------------
src/hotspot/share/gc/cms/jvmFlagConstraintsCMS.cpp
 153                           "a multiple of " INT32_FORMAT "\n",

HeapWordSize is of type int.  Use %d directly, rather than
INT32_FORMAT.

------------------------------------------------------------------------------
src/hotspot/share/runtime/flags/jvmFlagConstraintsCompiler.cpp
  97                         "between 0 and %d\n",
  98                         AllocatePrefetchDistance, 512);

Why is this hard-coding the 0 but not 512 in the string.  Or better,
make 512 a named constant and replace both uses here.

Or even better, why is this a constraint function at all?  Couldn't
this apparently platform-generic restriction just be a range
constraint, e.g. range(-1, 512)?

------------------------------------------------------------------------------
src/hotspot/share/runtime/flags/jvmFlagConstraintsCompiler.cpp
 147 JVMFlag::Error OnStackReplacePercentageConstraintFunc(intx value, bool verbose) {

The clauses being changed are only reachable via a narrowing
conversion, which has implementation-defined value.  Fixing that is
probably out of scope for this change.

------------------------------------------------------------------------------
src/hotspot/share/runtime/flags/jvmFlagConstraintsCompiler.cpp
 398                         "RTMTotalCountIncrRate (" INT32_FORMAT

RTMTotalCountIncrRate is of type int.  Use %d directly, rather than
INT32_FORMAT.

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

Have you checked for uses of printError() in code specific to platforms
not tested by Oracle?