Closed simonjwright closed 4 years ago
Hi @simonjwright ,
Do you know what is optimized exactly?
It would be best to fix this in the source code rather than the project file. You change will break builds for RISC-V which doesn't have semihosting.adb
.
Point taken.
Commit cf3d488 works for me. I did look at the ASM but I really didn’t know what to look for; guessed that forcing copying byte-by-byte might do the trick.
The title of this pull request is now OBE; and it includes a commit that you wouldn’t want. Assuming you’re OK with the change, how should I proceed?
Thanks @simonjwright , the last patch looks good.
You can add a revert commit on this PR for e7884d5 or start from a new PR.
Either way please rebase on master since I just merged a change in the continuous integration scripts.
You change will break builds for RISC-V which doesn't have
semihosting.adb
.
I don’t believe this is true. This GPR from my Cortex GNAT RTS specifies Switches
for at least 3 files that aren’t present in that RTS.
Closed in favour of #353
This problem didn’t show up with GNAT CE 2019, but it does with FSF GCC 9.1.0 & 10.1.0: the output of semihosting is garbled if semihosting.adb is compiled with -O3, OK if compiled with -O0.
What should have been
came out as
Setup: GDB 8.3 & 9.1 rsp, SEGGER J-Link Edu, SEGGER J-Link GDB Server V6.42 Command Line Version, macOS Catalina