Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [cdt-dev] DSF: register view problems

Hi Dimitry,
This bug hasn't been reported before, but it's something we've seen with the memmory view as well: The register view coalesces requests for individual register values into a single command for a range of registers. However, if GDB finds a problem with any of the registers it reports an error for the entire command, so all the register values appear blank in the view. I'm not sure how we should best address it. It could be argued that GDB should be fixed and report partial results, or QEMU should be fixed to support the missing registers, but neither of these would help your situation. What we could do in DSF-GDB is to remove the coalescing completely, or add logic to request values individually in case the coalesced command fails, etc. Please file a bug to track this issue, we should be able to at least do something simple in 6.0.1.


Dmitry Smirnov wrote:
I'm observing the following problem with the register view.
I'm debugging an ARM program, that is running on QEMU simulator.

QEMU supports GDB stub API for register name/value retrieving and changing.
For the ARM core, it supports registers r0-r15, cpsr. There is a gap
between ro-r15 and cpsr registers: these registers are not supported
by QEMU and responded as bad register number.
If Register view of CDT (in DSF debug configuration) starts displaying
the registers from the beginnig and I scroll down row by row, every
register is displayed as expected: correct values for r0-r15(pc) and
error message as a value for others, correct value for cpsr.

If the one of the absent register is visible on the Register view and
I press Refresh for this view, ALL visible register becomes invalid
("Failed to execute MI command" as a value) including those that exist
(registers 10-15 in the log below).
What is interesting here, is that if one of the existing and visible
register is selected in the view, its name and value are displayed
correctly in the lower part of the Register view (like register number
10 in the log below) but not in upper half.

Here is the log of the traces for this case.

Shouldn't DSF skip requesting registers with empty names?

497,864 847-data-list-register-names
497,880 847^done,register-names=["r0","r1","r2","r3","r4","r5","r6","r7","r8","r9","r10","r11","r12"\
497,880 (gdb)
497,895 848-data-list-register-values x 10 11 12 13 14 15 16 17
497,895 849-data-list-register-values N 10 11 12 13 14 15 16 17
497,911 848^error,msg="bad register number"
497,911 (gdb)
497,911 850-data-list-register-values x 10
497,926 851-data-list-register-values N 10
497,926 849^error,msg="bad register number"
497,926 (gdb)
497,926 850^done,register-values=[{number="10",value="0x0"}]
497,926 (gdb)
497,926 851^done,register-values=[{number="10",value="0"}]
497,926 (gdb)
497,973 852-data-list-register-values t 10
497,973 853-data-list-register-values o 10
497,973 854-data-list-register-values d 10
497,973 852^done,register-values=[{number="10",value="0"}]
497,973 (gdb)
497,973 853^done,register-values=[{number="10",value="0"}]
497,973 (gdb)
497,973 854^done,register-values=[{number="10",value="0"}]
497,973 (gdb)
cdt-dev mailing list

Back to the top