qapi/parser: Don't try to handle file errors

Fixes: f5d4361cda
Fixes: 52a474180a
Fixes: 46f49468c6

Remove the try/except block that handles file-opening errors in
QAPISchemaParser.__init__() and add one each to
QAPISchemaParser._include() and QAPISchema.__init__() respectively.

This simultaneously fixes the typing of info.fname (f5d4361cda), A
static typing violation in test-qapi (46f49468c6), and a regression of
an error message (52a474180a).

The short-ish version of what motivates this patch is:

- It's hard to write a good error message in the init method,
  because we need to determine the context of our caller to do so.
  It's easier to just let the caller write the message.
- We don't want to allow QAPISourceInfo(None, None, None) to exist. The
  typing introduced by commit f5d4361cda types the 'fname' field as
  (non-optional) str, which was premature until the removal of this
  construct.
- Errors made using such an object are currently incorrect (since
  52a474180a)
- It's not technically a semantic error if we cannot open the schema.
- There are various typing constraints that make mixing these two cases
  undesirable for a single special case.
- test-qapi's code handling an fname of 'None' is now dead, drop it.
  Additionally, Not all QAPIError objects have an 'info' field (since
  46f49468), so deleting this stanza corrects a typing oversight in
  test-qapi introduced by that commit.

Other considerations:

- open() is moved to a 'with' block to ensure file pointers are
  cleaned up deterministically.
- Python 3.3 deprecated IOError and made it a synonym for OSError.
  Avoid the misleading perception these exception handlers are
  narrower than they really are.

The long version:

The error message here is incorrect (since commit 52a474180a):

> python3 qapi-gen.py 'fake.json'
qapi-gen.py: qapi-gen.py: can't read schema file 'fake.json': No such file or directory

In pursuing it, we find that QAPISourceInfo has a special accommodation
for when there's no filename. Meanwhile, the intent when QAPISourceInfo
was typed (f5d4361cda) was non-optional 'str'. This usage was
overlooked.

To remove this, I'd want to avoid having a "fake" QAPISourceInfo
object. I also don't want to explicitly begin accommodating
QAPISourceInfo itself being None, because we actually want to eventually
prove that this can never happen -- We don't want to confuse "The file
isn't open yet" with "This error stems from a definition that wasn't
defined in any file".

(An earlier series tried to create a dummy info object, but it was tough
to prove in review that it worked correctly without creating new
regressions. This patch avoids that distraction. We would like to first
prove that we never raise QAPISemError for any built-in object before we
add "special" info objects. We aren't ready to do that yet.)

So, which way out of the labyrinth?

Here's one way: Don't try to handle errors at a level with "mixed"
semantic contexts; i.e. don't mix inclusion errors (should report a
source line where the include was triggered) and command line errors
(where we specified a file we couldn't read).

Remove the error handling from the initializer of the parser. Pythonic!
Now it's the caller's job to figure out what to do about it. Handle the
error in QAPISchemaParser._include() instead, where we can write a
targeted error message where we are guaranteed to have an 'info' context
to report with.

The root level error can similarly move to QAPISchema.__init__(), where
we know we'll never have an info context to report with, so we use a
more abstract error type.

Now the error looks sensible again:

> python3 qapi-gen.py 'fake.json'
qapi-gen.py: can't read schema file 'fake.json': No such file or directory

With these error cases separated, QAPISourceInfo can be solidified as
never having placeholder arguments that violate our desired types. Clean
up test-qapi along similar lines.

Signed-off-by: John Snow <jsnow@redhat.com>
Message-Id: <20210519183951.3946870-2-jsnow@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
4 files changed