Possibly serious bug in a basic ability (interface to Maxima).

[Emmanuel Charpentier]

I may be wrong, but a bit of exploration of Trac#28431, which seemed fairly innocuous, may demonstrate a basic failure in our interface to Maxima, which is, IMHO, more severe than a bug in Maxima...

I also checkef that's not a side effect of the domain:complex setup...

A look by someone who knows what he's doing in this region of Sage may be worthy...

[rjf]

I just tried the problem in Maxima 5.40.0 ; the result is correct as I assume

Emmanuel also discovered.  The "fix" suggested by the poster of Trac#28431

(to divert some class of integral problems to sympy) does not strike me

as plausible.

Maybe something about Sage parsing of a minus sign??

RJF

[rjf]

I previously had not looked at the further discussion in that report,

but the discussion of branches of the log function and the Sage

"simplification" shows it is not a parsing problem.

Since log() is in general multivalued, any answer that

ignores this possibility may fall into a trap.  Setting a domain

to something or other does not necessarily avoid the trap.

An example that is simpler and perhaps reminiscent of this

kind of error is to assert that   sqrt(x^2) is x  or abs(x), when

there are (except if x=0)  TWO square roots,  x, -x.

Even if you know that x is positive,  there are still two

square roots. sqrt(9) has 2 values.  Unless you want

to define sqrt as something else.  Do you?  How about log()

in the current example..

RJF