The main problem was that on MacOS, expr didn't like the regexp
operator "+" in, say, "[0=9]+". This is easily fixed by using the
equivalent "[0-9][0-9]*".
We also added some more checks to make sure that makeinfo_version is
not empty and that makeinfo_major and makeinfo_minor are not empty
either. If any of these are empty, produce an error message.
Tested this on MacOS and verified that the warning messages are
produced (by using a faulty regexp for expr).
BOTH-SIDE: fix case where limit from below is und, from above is bounded
(The opposite case where the limit from below is bounded and the limit
from above is und was not problematic.)
Commit df1748a8 had a bug which caused this specific case to yield ind
instead of und:
(%i1) limit(exp(1/x)*sin(1/x),x,0); /* correct */
(%o1) und
(%i2) limit(exp(-1/x)*sin(1/x),x,0); /* incorrect */
(%o2) ind
I misread the beginning of BOTH-SIDE and I thought the case where a
one-sided limit yielded und was taken care of, but that was only true
for one direction.
There have been no tests in the test suite where the limit from one
direction is und and from the other direction is something bounded.
No problems with the test suite, share test suite or rtest_translator.
New tests have been added to rtest_limit.
318dd5a...
by
Barton Willis <email address hidden>
o Marking 56 & 57 of rtest_limit_extra as known failures. The expected output was incorrect and only recently fixed.
813932e...
by
Barton Willis <email address hidden>
First, fix typo: "{$makeinfo_old}" should be "${makeinfo_old}". That
was preventing us from signaling an error.
Second, make sure if --disable-build-docs is set, we don't try to test
for the makeinfo version. Do this by moving all of the code inside
the test for enable_build_docs being true.
Added `makeinfo_old` variable that is set to true if makeinfo is too
old. Update that appropriately, and print out an error or message
based on the value of `makeinfo_old`.