-Wall doesn't include all warnings

[guest]

Why is the monomorphism restricion warning different from other warning? I think it should have a -W flag.

Furthermore, having this flag on by default breaks package builds that use -Werror.

-- Lennart

[igloo]

(Can you please not close bugs without comment or anonymously? Thanks)

On the name of the flag, -fwarn rather than -W seems to be the naming convention.

However, on having the warning on by default:

We've seen a number of failures in the HEAD+extralibs build because of this warning, e.g. a warning about index in the below snippet from regex-compat:

else let index = (read (head bgroups)) - 1
     in
     if index == -1
        then match
        else groups !! index

Here, and I imagine everywhere (non-contrived) that compilation succeeds, monomorphism is exactly what we want. As I believe it is a good policy to keep code -Wall clean, I personally would not like this warning to be enabled even by -Wall. Debatably it would be good style to add a type signature for index, though.

The warning may be useful for people investigating removing or replacing the MR, but they can enable it themselves.

It may also be useful to show any such warnings when compilation fails, but that feels like a bit of a hack and may result in a flood of unrelated warnings when a single error occurs.

I've turned the warning off for now, so at least it won't break the nightly builds.

[simonpj]

Let me observe that

• Many newbie complaints are due to silent application of the MR
• It can always be suppressed by adding a type signature

Perhaps one could make a case to allowing silent MR for nested defns, like the one above.

In any case, I don't feel strongly enough to reopen this. For now it can stay off by default.

Simon

[simonmar]

-Wall should enable all warnings, otherwise it's badly named. So if we have -fwarn-monomorphism-restriction at all, it should be in -Wall, and possibly also in -W. I personally think it's useful and we should have it, but it should not be enabled by default.

If you use -Wall -Werror, you should expect your code to break whenever a new warning is added. Perhaps we should emit a warning for this option combination :-)

Incedentally, the following warnings are not in -Wall (I believe we should fix this):

• -fwarn-simple-patterns
• -fwarn-tabs
• -fwarn-incomplete-record-updates

[igloo]

gcc, from which I assume we get the name -Wall, says in its info page:

The following `-W...' options are not implied by `-Wall'.  Some of
them warn about constructions that users generally do not consider
questionable, but which occasionally you might wish to check for;
others warn about constructions that are necessary or hard to avoid in
some cases, and there is no simple way to modify the code to suppress
the warning.

While it is poorly named, I think that the above list plus -fwarn-monomorphism-restriction fall into the same category.

[simonmar]

Ok, I'm convinced. -Wall would be rather less useful if it really included all warnings. I've updated the documentation to match the current behaviour.

I'm ambivalent about whether -fwarn-monomorphism-restriction should be in -Wall; Simon PJ believes it should be. That discussion can be moved to the mailing list, however.