Doesn't look like fir1 has changed substantially between versions, but fir2 has. Here's a comparison between octave-signal_1.1.3 and 1.2.2 in Debian (1.2.0 is quite similar).

[~/Downloads/signal/inst]$ diff fir1.m /usr/share/octave/packages/signal-1.2.2/fir1.m
143,145c143,145
< %!assert(fir1(2, .5, 'low', @hanning, 'scale'), [0 1 0]');
< %!assert(fir1(2, .5, 'low', "hanning", 'scale'), [0 1 0]');
< %!assert(fir1(2, .5, 'low', hanning(3), 'scale'), [0 1 0]');
---

> %!assert(fir1(2, .5, 'low', @hanning, 'scale'), [0 1 0]);
> %!assert(fir1(2, .5, 'low', "hanning", 'scale'), [0 1 0]);
> %!assert(fir1(2, .5, 'low', hanning(3), 'scale'), [0 1 0]);

[~/Downloads/signal/inst]$ diff fir2.m /usr/share/octave/packages/signal-1.2.2/fir2.m | wc -l
89


Furthermore, simple substitution of the old fir2 into a bare directory executes without error. I haven't yet checked that the output is correct. Qualitative checks with my own code and with freqz both suggest that it may be.

[~/octavebug]$ ls
fir2.m
[~/octavebug]$ octave
GNU Octave, version 3.6.4
...
octave:1> size(fir1(10000,0.01))
ans =

   10001       1

octave:2>

b = fir1(8000,0.01) returns error message, "error : fir2 : grid size must be greater than half the filter order. "

Using Octave UPM R8.1, the fir1() function in 'Signal' package