On Sun, Jun 7, 2015 at 3:51 AM, Peter Maydell <peter.mayd...@linaro.org> wrote: > On 7 June 2015 at 09:49, Peter Crosthwaite <peter.crosthwa...@xilinx.com> > wrote: >> OK I am at the bottom of it. The case statement only handles the base >> arch and the host arch not the actual target arch. > > Aha. > >> This means the >> "arm)" case is all that is called for aarch64 target. the "aarch64)" >> case in existing code is presumably to handle an AArch64 host. So the >> current code will produce the minimal working configury. My patch will >> still work, but will unneedingly add AArch32 disas to AA64 host. > > Presumably also the current code will unnecessarily add an A64 > disassembler for all AArch32 host builds. > >> A solution would be to patch that loop to go through all of the base >> arch, target arch and host arch. Anyone know of a reason not to do so? > > Not offhand, but then I didn't figure out how the current code > worked when I looked at it :-) We'd get duplicate lines in > the config files, but we already have those when host==target. >
Brute force it and sort -u? sort is considered a coreutil by at least my manpages so I wonder if we can assume it is universally available. Regards, Peter > -- PMM >