[v3,0/3] RISC-V/Linux `gdbserver' support and associated fixes

Message ID alpine.LFD.2.21.2001311115550.14118@redsun52.ssa.fujisawa.hgst.com
Headers show
  • RISC-V/Linux `gdbserver' support and associated fixes
Related show


Maciej W. Rozycki Jan. 31, 2020, 12:11 p.m.

 This is v3 of my RISC-V/Linux `gdbserver' support proposal.

 Beyond the issues discussed with v2 I have now also slightly optimised 
regcache supply/collect handlers to avoid doing a costly variable 
multiplication in a loop, and added buffer offset precalculations so as to 
avoid excessive line wrapping and hopefully making code more readable.

 Also I have noticed missing Python and ncurses development libraries 
limiting testing in my native setup.  With these installed native test 
results improved a little, as follows:

		=== gdb Summary ===

# of expected passes		61354
# of unexpected failures	1636
# of unexpected successes	1
# of expected failures		58
# of unknown successes		3
# of known failures		85
# of unresolved testcases	113
# of untested testcases		160
# of unsupported tests		323

however a worrying regression has appeared:

FAIL: gdb.base/return-nodebug.exp: float: full width of the returned result

This has turned out not to be related to this patch series however and 
triggers reliably now in my setup regardless of whether this patch set has 
been applied or not.  Instead it is an outcome of GDB failing to NaN-box 
data of the `float' type when assigning it to a hardware register.

 I have filed PR tdep/25489 to track it as I may not be able to look into 
it right away.

 Except as noted above here have been no regressions in native 
`riscv64-linux-gnu' testing and remote `gdbserver' test results are the 
same as previously (barring the usual gdb.threads/ fluctuations).

 As usually see individual changes for details.



Maciej W. Rozycki Jan. 31, 2020, 12:49 p.m. | #1
On Fri, 31 Jan 2020, Maciej W. Rozycki wrote:

>  This is v3 of my RISC-V/Linux `gdbserver' support proposal.

 Apologies for the series being broken, there appears to be something here 
in the e-mail setup that causes `In-Reply-To:' and `References:' headers 
to be randomly dropped sometimes and I haven't figured out what the exact 
circumstances are.

 In any case the changes themselves are: