diff options
author | Thiago Jung Bauermann <thiago.bauermann@linaro.org> | 2024-08-25 20:34:13 -0300 |
---|---|---|
committer | Thiago Jung Bauermann <thiago.bauermann@linaro.org> | 2024-09-05 01:02:51 -0300 |
commit | 43af2e08dc0af7796b557d14f13317c0c24f948a (patch) | |
tree | d43173382ad9a6a56a834a32f8b0f003a06b2143 /bfd | |
parent | 88327cbcd02807149450b42ded5949ae6ab845b7 (diff) | |
download | gdb-43af2e08dc0af7796b557d14f13317c0c24f948a.zip gdb-43af2e08dc0af7796b557d14f13317c0c24f948a.tar.gz gdb-43af2e08dc0af7796b557d14f13317c0c24f948a.tar.bz2 |
gdbserver: aarch64: Fix expedited registers list
Since this commit:
commit a8651ef51822f91ec86d0d5caffbf2e50b174c23
CommitDate: Fri Jun 14 14:47:38 2024 +0100
gdb/aarch64: prevent crash from in process agent
gdbserver isn't sending expedited registers with its stop reply packets
anymore. The problem is with how the constructor of the
expedited_registers std::vector is called:
The intent of the expedited_registers initialization in
aarch64_linux_read_description is to create a vector with capacity for 6
elements, but that's not how the std::vector constructor works.
Instead it creates a vector pre-populated with 6 elements initialized
with the default value for the type of the elements, and thus the first
6 elements are null pointers. The actual expedited registers are added
starting at the 7th element.
This causes init_target_desc to consider that the expedite_regs list is
empty, since it stops checking at the first nullptr element. The end
result is that gdbserver doesn't send any expedited registers to GDB in
its stop replies.
Fix by not specifying an element count when declaring the vector.
Tested for regressions on aarch64-linux-gnu native-extended-remote.
Approved-By: Andrew Burgess <aburgess@redhat.com>
Diffstat (limited to 'bfd')
0 files changed, 0 insertions, 0 deletions