New semicolon handling in specs uncovered questionable error text

New semicolon handling in specs uncovered questionable error text

[Loren James Rittle]

Note the semicolon usage in the spec fragments below.  Did a recent
mainline patch break this usage?  How may I now quote said semicolon
usage in that error reporting context?  I am sorry to bother the list
with this trivial stuff but I see no gcc/ChangeLog entry or new
commentary on spec formats to support the change to gcc/gcc.c that
would cause the failure mode (I do see the comment on the new spec
form that uses semicolons but I don't see how %e text would match it).
I suspect it is -r1.351 of gcc/gcc.c (but I thought I already resolved
the issues with that merge into mainline thus perhaps it is a latent issue).

gcc/config/alpha/freebsd.h:  %{p:%e`-p' not supported; use `-pg' and gprof(1)} \
gcc/config/ia64/freebsd.h: "%{p:%e`-p' not supported; use `-pg' and gprof(1)}  \
gcc/config/sparc/freebsd.h:  %{p:%e`-p' not supported; use `-pg' and gprof(1)} \
gcc/config/i386/freebsd-aout.h:  "%{p:%e`-p' not supported; use `-pg' [...]

Now, is there a better way to handle this situation than just adapting
the error message?  I checked one random RH8 GNU/Linux box and I note
that it too has no prof(1) yet gcc there doesn't bother to inform the
user that -p is not supported.  Perhaps, I should convert all FreeBSD
targets to include the following notice instead:

 %{p:%nconsider using `-pg' with gprof(1)}

Or, should the entire line be axed since it appears that -p actually
produces a file that is readable by gprof(1)?  It is just missing
some/most of the possible information required for full gprof reports.

Regards,
Loren

Re: New semicolon handling in specs uncovered questionable error text

[Zack Weinberg]

Loren James Rittle <rittle@latour.rsch.comm.mot.com> writes:

> Note the semicolon usage in the spec fragments below.  Did a recent
> mainline patch break this usage?

The %{a:b;c:d} N-way choice patch.

We could make it possible to quote this, but have you considered
commas?

zw

Re: New semicolon handling in specs uncovered questionable error text

[Loren James Rittle]

> The %{a:b;c:d} N-way choice patch.

> We could make it possible to quote this, but have you considered
> commas?

Yes, a comma works; quoting not required.  Based on local testing, I
am going to patch it to weaken the error to a warning (under separate
cover to gcc-patches).

Thanks,
Loren