Hello Guinevere,
Guinevere Larsen via linaro-toolchain linaro-toolchain@lists.linaro.org writes:
It seems that the system running the armv8l has a pretty old pascal compiler. I checked the log and it seems that, with that compiler version, using fpc -g will generate stabs information, rather than DWARF information.
So in one sense, these regressions are expected, but more importantly, you should take a look at updating the armv8l machine so that pascal tests can continue working
Thank you for checking! We use the fpc compiler from Ubuntu 24.04 in all our CI machines, which is at version 3.2.2. It's the latest version according to the fpc website¹ (even though there's a 3.2.3 tag in the repository).
The problem is actually that on armv8l-linux the -g option defaults to stabs, even though on aarch64-linux and x86_64-linux it defaults to DWARF. I opened an issue upstream about it².
FWIW, if I hard-code fpc_compile in the GDB testsuite to use -gw to generate DWARF, then I get these results on armv8l-linux, which are the same as the ones on aarch64-linux and x86_64-linux:
=== gdb Summary ===
# of expected passes 169 # of known failures 1