Anutosh Bhat

I would try to approach this soon ! EDIT: This error seems to be a result of `calculate_series` method in `gruntz.py` , which is returning different results for different arrangements...

The error lies mainly in `add._eval_as_leading_term()` and `power._eval_nseries()` ``` >>> k, n = symbols('k, n', positive=True, integer=True) >>> -n**(-k - 1) + (1 + 1/n)**(k + 1) -n**(-k - 1)...

But on the other hand I suppose `(1 + 1/n)**(k + 1).)_eval_nseries(n, -k-1, None)` call shouldn't be made . This arises from this block( line 1052 - 1055) where min...

The only thing to be fixed now is series for `(1 + 1/n)**k + (1 + 1/n)**k/n - 1/(n*n**k)` ``` >>> expr = (1 + 1/n)**k + (1 + 1/n)**k/n...

Hello @oscarbenjamin thanks for opening this issue . It is a known issue but there wasn't any open issue for this . Actually I have a Pr ( #22555) trying...

It seems this would also solve the series errors from #7224 ( the issue also needs fixes in the solvers module) but would solve the series issues reported there ....

After making the changes , there were 2 errors from `test_wester.py` which weren't working as expected. The latest commit fixes 1 of them . The error was ``` >>limit(log(x)/(log(x) +...

> How does the addition of AccumBounds to `_sign` help with `gruntz.sign`? Not a hard requirement, I put that just in case , thinking it might have a role to...

I guess we need to have a block to handle infinites better . As I was trying out asymptotic limits along directions other than just `+1/-1`..... I ended up trying...

Limits for Shi function at `+oo/-oo` were giving wrong outputs due to a wrongly placed check in `Shi._eval_as_leading_term()`, the latest commit also fixes that . ``` >>> # On master...