Fix formatting error on 32-bits targets

[dkm]

Printing size_t as [unsigned] long (%ld or %lu) raises warnings on 32-bits
targets. As the GCC pretty printer doesn't have the equivalent of libc's %z/%zu,
fix all formats to use unsigned long and cast values.

refs #1229

Signed-off-by: Marc Poulhiès [email protected]
Co-authored-by: Rainer Orth [email protected]

[tschwinge]

This raises a lot of `log_warnings` (#1026) diagnostics -- so that's working, good! ;-P Instead of plain `%z`, you want `%zu` for `size_t`, or `%zi`/`%zd` for `ssize_t`. Though, I see almost no `z` length modifier usage in GCC. Are there compatibility concerns? (Still with GCC requiring C++11 nowadays?) (For avoidance of doubt: I'm all in favor of using `z` if permissible.)

[dkm]

Woops of course should have used zu :)
I'll dig a bit to see if there's more to its absence than simply force of habit :)

[CohenArthur]

I believe the z modifier was added for C99 so we should be good on the compatibility part?

The man page for printf states the following:

glibc 2.1 adds length modifiers hh, j, t, and z and conversion characters a and A.

and 2.1 was released waaaaay before GCC 4.8 / C++11, so I assume we should be in the clear?

I think this is definitely something that should go in

[CohenArthur]

LGTM :)

[dkm]

Reverted back to cast as %z spec not supported by GCC pretty printers

[CohenArthur]

Ah well that's a bit sad :) So no one is interested in patching pp_format and submitting it? :P

[dkm]

Ah well that's a bit sad :) So no one is interested in patching pp_format and submitting it? :P

Thought about that, but that was before looking at the code :) I'm not sure that's really worth the effort...

[dafaust]

LGTM. Bit of a shame about %z but I won't blame anyone for not wanting to patch that :P

[CohenArthur]

bors r+

[bors]

Build succeeded:

• build-and-check

[dkm]

Thanks for merging it, forgot about it sorry !