fix Issue 12979 - Nothrow violation error is hidden by inline assembler by MartinNowak · Pull Request #4033 · dlang/dmd

MartinNowak · 2014-09-29T09:10:21Z

deprecate using unattributed asm in attributed functions
add explicit attributes on asm statments (unverified by compiler)

asm pure nothrow @nogc @trusted {
//...
}
added new CompoundAsmStatement because the parser
already splits asm blocks into many AsmStatements

MartinNowak · 2014-09-29T09:31:09Z

This triggers many warnings in druntime about missing returns, which probably happens because of the added blockExit check, so I'll have to revise the pull.

WalterBright · 2014-09-29T09:56:29Z

Adding new grammar will also require documentation pulls.

ibuclaw · 2014-09-29T09:57:27Z

cool

9rnsr · 2014-09-29T11:14:12Z

test/fail_compilation/iasm_attributes.d

For the language consistency, I think we should also support @trusted, @nogc and pure attributes on iasm statement, and we need to change a bare iasm as impure, unsafe, gc-able, and throwable by default.

I think so too.

MartinNowak · 2014-09-29T21:47:34Z

There is also a grammar pull dlang/dlang.org#667.

MartinNowak · 2014-09-29T23:01:07Z

Requires dlang/druntime#974 first.

WalterBright · 2014-09-30T06:49:26Z

How can this be marked as [enhancement] in the PR list?

yebblies · 2014-09-30T06:56:31Z

How can this be marked as [enhancement] in the PR list?

On this page, go to the top-right corner, just to the right of the PR description, and click "Labels" then click "Enhancement".
On the pulls page check the box next to this pull then select the "Label" dropdown and click "Enhancement".

WalterBright · 2014-09-30T07:18:06Z

Thanks, @yebblies, though could someone else please do this as github doesn't like my laptop.

9rnsr · 2014-09-30T13:04:32Z

This enhancement will fill the semantic gap between inline assembler and function attributes.

9rnsr · 2014-09-30T13:04:38Z

Looks good.

9rnsr · 2014-09-30T13:19:49Z

src/statement.h

In dmd code CompoundDeclarationStatement already exists, so the name CompoundAsmStatement might be better?

I was actually thinking about AsmStatement and renaming the old AsmStatement to AsmInstruction. That would be far less confusing.

This is because AsmStatement::semantic splits them up into one-per-instruction blocks right? Can't we just defer that split to the glue layer somehow?

In that case, I should probably rename GDC's as ExtAsmTemplate. ;)

This is because AsmStatement::semantic splits them up into one-per-instruction blocks right?

The parser already does that. I think the main reason for this are LabelStatements which can be accessed from outside asm blocks.

Can't we just defer that split to the glue layer somehow?

Maybe, but this isn't part of this pull request.

Maybe, but this isn't part of this pull request.

Fair enough.

9rnsr · 2014-10-02T14:25:13Z

@MartinNowak Druntime PR was merged, so this PR should be ready to merge.
But, auto-tester shows an error in fail_compilation/deprecate12979a.d. Let's fix it.

If others have no complaint about this enhancement, I'll merge this in this weekend (UTC 10/5 0:00).

MartinNowak · 2014-10-02T16:59:36Z

But, auto-tester shows an error in fail_compilation/deprecate12979a.d.

done

- deprecate using unattributed asm in attributed functions - add explicit attributes on asm statments (unverified by compiler) asm pure nothrow @nogc @trusted { //... } - added new CompoundAsmStatement because the parser already splits asm blocks into many AsmStatements

WalterBright · 2014-10-04T20:50:20Z

src/statement.c

asm statement, not asm block, as we should consistently refer to things with the same phrase. I also don't like the double negative, and would prefer something along the lines of:

asm statement is assumed to throw - mark it with 'nothrow' attribute if it does not

asm statement, not asm block, as we should consistently refer to things with the same phrase.

I mixed up the wording because AsmStatement in the compiler is actually a single asm instruction.

9rnsr · 2014-10-05T00:00:03Z

@WalterBright You would have no objection about the enhancement feature itself, right?.

@MartinNowak I'll defer the merging until your reaction comes back.

MartinNowak · 2014-10-05T19:10:35Z

Updated the wording of the error messages.

dnadlinger · 2014-10-05T19:24:40Z

@redstar, @AlexeyProkhin: When this is in, we can probably get rid of our AsmBlockStatement (LDC keeps asm blocks as one for different reasons).

9rnsr · 2014-10-06T13:05:06Z

LGTM.

9rnsr · 2014-10-06T13:05:27Z

Auto-merge toggled on

fix Issue 12979 - Nothrow violation error is hidden by inline assembler

9rnsr · 2014-10-08T15:11:39Z

From a practical reason, I'd propose an additional enhancement https://issues.dlang.org/show_bug.cgi?id=13589

MartinNowak force-pushed the fix12979 branch from 0000000 to 627661c Compare September 29, 2014 09:10

MartinNowak mentioned this pull request Sep 29, 2014

Remove error for asm statements in a nothrow function #4030

Closed

9rnsr reviewed Sep 29, 2014
View reviewed changes

MartinNowak force-pushed the fix12979 branch from 627661c to 8f88c05 Compare September 29, 2014 21:30

MartinNowak force-pushed the fix12979 branch from 8f88c05 to 8827e2e Compare September 29, 2014 21:50

MartinNowak changed the title ~~fix Issue 12979 - Nothrow violation error is hidden by inline assembler~~ parse postfix attributes on asm blocks Sep 29, 2014

MartinNowak force-pushed the fix12979 branch from f7181a8 to 8821a97 Compare September 29, 2014 22:27

ibuclaw added the Severity:Enhancement label Sep 30, 2014

9rnsr reviewed Sep 30, 2014
View reviewed changes

monarchdodra mentioned this pull request Sep 30, 2014

attributes for inline asm dlang/phobos#2572

Merged

MartinNowak force-pushed the fix12979 branch 2 times, most recently from e9cdb3e to faa3bfc Compare September 30, 2014 23:14

MartinNowak changed the title ~~parse postfix attributes on asm blocks~~ fix Issue 12979 - Nothrow violation error is hidden by inline assembler Sep 30, 2014

MartinNowak force-pushed the fix12979 branch from faa3bfc to 4b4b08e Compare October 2, 2014 16:58

MartinNowak force-pushed the fix12979 branch from 4b4b08e to c8fd9db Compare October 2, 2014 17:07

MartinNowak force-pushed the fix12979 branch from c8fd9db to 31b8ff1 Compare October 3, 2014 19:14

WalterBright reviewed Oct 4, 2014
View reviewed changes

wording of error messages

ea3237b

MartinNowak force-pushed the fix12979 branch from 5011b89 to ea3237b Compare October 6, 2014 11:00

9rnsr added a commit that referenced this pull request Oct 6, 2014

Merge pull request #4033 from MartinNowak/fix12979

a1ec5c7

fix Issue 12979 - Nothrow violation error is hidden by inline assembler

9rnsr merged commit a1ec5c7 into dlang:master Oct 6, 2014

ibuclaw mentioned this pull request Oct 6, 2014

Update frontend to 2.066 D-Programming-GDC/gdc#87

Merged

MartinNowak deleted the fix12979 branch October 6, 2014 17:46

This was referenced Oct 29, 2014

C-style array fixes are incorrectly applied to assembly dlang-community/dfix#6

Closed

Add support for attributes on ASM statements dlang-community/libdparse#19

Closed

Uh oh!

Comments

Conversation

MartinNowak commented Sep 29, 2014

Uh oh!

MartinNowak commented Sep 29, 2014

Uh oh!

WalterBright commented Sep 29, 2014

Uh oh!

ibuclaw commented Sep 29, 2014

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

MartinNowak commented Sep 29, 2014

Uh oh!

MartinNowak commented Sep 29, 2014

Uh oh!

WalterBright commented Sep 30, 2014

Uh oh!

yebblies commented Sep 30, 2014

Uh oh!

WalterBright commented Sep 30, 2014

Uh oh!

9rnsr commented Sep 30, 2014

Uh oh!

9rnsr commented Sep 30, 2014

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

9rnsr commented Oct 2, 2014

Uh oh!

MartinNowak commented Oct 2, 2014

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

9rnsr commented Oct 5, 2014

Uh oh!

MartinNowak commented Oct 5, 2014

Uh oh!

dnadlinger commented Oct 5, 2014

Uh oh!

9rnsr commented Oct 6, 2014

Uh oh!

9rnsr commented Oct 6, 2014

Uh oh!

9rnsr commented Oct 8, 2014

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants