__futures__ parsing can be simplified using complete source locations for error detection. Also the error's offset is off by one sometimes.

[iritkatriel]

The check for whether there are __future__ imports which are not in the beginning of the script appears in two places:

in future.c, future_parse, the ff_lineno is stored (last lineno of a __future__ import) and there is a check for bad __future__ imports on the same line .

Then in compile.c, compiler_from_import(), any import from __future__ which is in a line after c->c_future->ff_lineno is rejected.

If we replace ff_lineno by ff_location (the complete location information) then the check in compile.c can use the column info to detect the same-line case, and we no longer need to check for this in future.c. The code of future_parse will become much simpler.

There is also a bug in the future.c exception - the offset is off by 1, and this results in incorrect hilighting of the bad import:

python.exe -c "import test.badsyntax_future7.py"
Traceback (most recent call last):
  File "<string>", line 1, in <module>
  File "/Users/iritkatriel/src/cpython/Lib/test/badsyntax_future7.py", line 3
    from __future__ import nested_scopes; import string; from __future__ import \
                                                        ^
SyntaxError: from __future__ imports must occur at the beginning of the file

My patch will result in this output instead:

Traceback (most recent call last):
  File "<string>", line 1, in <module>
  File "/Users/iritkatriel/src/cpython-654/Lib/test/badsyntax_future7.py", line 3
    from __future__ import nested_scopes; import string; from __future__ import \
                                                         ^^^^^^^^^^^^^^^^^^^^^^^^
SyntaxError: from __future__ imports must occur at the beginning of the file

• PR: gh-98811: use full source location to simplify __future__ imports error checking. This also fixes an incorrect error offset. #98812