Andy Porter

fparser2 accepts an array with defined extent and the ALLOCATABLE attribute

1

This breaks C510 in the spec (https://wg5-fortran.org/N1601-N1650/N1601.pdf#sr.577).

Although we recently (#170) moved to allowing lists to contain 1 item, it seems this change hasn't carried through to the handling of routine arguments. When parsing: subroutine tmp(a, b)...

Using index() on the `BlockBase.content` list does not work

8

I've fallen over this twice now. I think the problem occurs because `Base` (and therefore `BlockBase`) inherit from `ComparableMixin` which re-defines the comparison operators. As a result, using `my_node.content.index(other_node)` fails...

Use argparse instead of optparse

4

In `script_options.py` fparser uses the `optparse` module for command-line argument processing. This module has now been deprecated in favour of `argparse`.

(Closes #2515) fix bugs in symbol-table merge

6

As it says. Primarily this is about making sure we don't rename symbols that are being imported into scope from elsewhere. Broken out of #924 as that is getting big.

Andy C. has reported that the RTD theme is not good for people who are colour blind etc. The Met Office has switched to using the 'pydata' theme for their...

The aforementioned intrinsics are "archaic" (according to the gnu webpages) versions of LOG, LOG10 and INT. Also, the `FLOAT` intrinsic is not currently listed as being available on GPU (L797...

Improve IncreaseRankLoopArray for cases with multiple uses of the variable

2

Currently PSyclone will refuse to parallelise a loop such as: DO IP=1,NP ! COSU = COS(USDIR(IP)) SINU = SIN(USDIR(IP)) ! DO ITH=1, NTH DIRF(ITH) = MAX ( 0. , (ECOS(ITH)*COSU+ESIN(ITH)*SINU)...

The WAVEWATCH 3 mini-app uses the Fortran `forall` construct. We need to translate this to a loop in the PSyIR (possibly recording that it was a forall originally in an...