Bug#860739: libghc-xcb-types-dev: cannot parse xcb-proto 1.12

[Iain R. Learmonth]

Hi,

On 07/19/2017 07:43 AM, Iain R. Learmonth wrote:

It would look that this is no longer a bug.


Nope, this is a bug. Was looking at the wrong test suite. :/

https://tests.reproducible-builds.org/debian/rb-pkg/unstable/amd64/cairocffi.html

For some reason this is only affecting python3 though...

Thanks,
Iain.

Bug#860739: libghc-xcb-types-dev: cannot parse xcb-proto 1.12

[Iain R. Learmonth]

Hi,

On 07/19/2017 04:49 AM, Clint Adams wrote:

On Sun, Apr 23, 2017 at 01:22:33PM +0100, James Cowgill wrote:

It looks like the xcffib dependency from cairocffi is optional, so this
would be a much easier solution. Otherwise we have to backport tons of
stuff or use upstream 0.5.1.


So does this mean this is no longer a bug?


xcffib is now building the latest upstream in unstable successfully and 
the test suite is passing, the test suite for qtile (which uses xcffib) 
is also passing.


It would look that this is no longer a bug.

Thanks,
Iain.

Bug#860739: libghc-xcb-types-dev: cannot parse xcb-proto 1.12

[Clint Adams]

On Sun, Apr 23, 2017 at 01:22:33PM +0100, James Cowgill wrote:
> It looks like the xcffib dependency from cairocffi is optional, so this
> would be a much easier solution. Otherwise we have to backport tons of
> stuff or use upstream 0.5.1.

So does this mean this is no longer a bug?

Bug#860739: libghc-xcb-types-dev: cannot parse xcb-proto 1.12

[James Cowgill]

Hi,

On 22/04/17 20:48, Niels Thykier wrote:
> On Sat, 22 Apr 2017 18:13:50 +0100 James Cowgill 
>> On Wed, 19 Apr 2017 16:05:12 +0100 James Cowgill  wrote:
>>> Package: libghc-xcb-types-dev
>>> Version: 0.7.1-4
>>> Severity: grave
>>> Tags: sid stretch
>>> Control: block 839314 by -1
>>>
>>> Hi,
>>>
>>> Currently haskell-xcb-types cannot parse xcb-proto 1.12. While strictly
>>> speaking you don't have to use this package to just parse the files from
>>> xcb-proto, it is the package's main purpose and (afaik) only use in the
>>> archive.
>>
>> [...]
>>
>> Relevant xcb-types commits:
>> https://github.com/aslatter/xcb-types/commit/0991f1d61b92371e9af51ab0fa3699d7c32e2b65
>> https://github.com/aslatter/xcb-types/commit/a86e578f0860670f3d43fa2d414e93a60aa72e2d
>>
>> Commit which would need backporting to xcffib:
>> https://github.com/tych0/xcffib/commit/6e6646b7d5eec253225579e1cbb40acf380648b4
>>
>> (xcffib would then need to depend on xcb-types >= 0.8)

More investigation!

After doing what I said above, xcb-types and xcffib build but the xcffib
testsuite fails[1]. It seems that xcb-proto 1.12 now uses the "switch"
element in a few places it didn't before and xcffib wasn't handling that
properly. There are a lot of commits in 0.5.0 which try to fix this - if
I use upstream xcffib it works.

> Hi James,
> 
> Thanks for the update and for looking into this issue.  I am glad to see
> there is progress on this issue.
> 
> I hope we can get a timely resolution on this bug and #839314.  If not,
> I will request that cairocffi drops its Build-Dependency and then remove
> xcffib plus haskell-xcb-types from testing.

It looks like the xcffib dependency from cairocffi is optional, so this
would be a much easier solution. Otherwise we have to backport tons of
stuff or use upstream 0.5.1.

[1] https://github.com/tych0/xcffib/pull/83

James



signature.asc
Description: OpenPGP digital signature

Bug#860739: libghc-xcb-types-dev: cannot parse xcb-proto 1.12

[Iain R. Learmonth]

Hi,

On Sat, Apr 22, 2017 at 07:48:00PM +, Niels Thykier wrote:
> Thanks for the update and for looking into this issue.  I am glad to see
> there is progress on this issue.

+1. Sorry I've not been able to look at this.

> I hope we can get a timely resolution on this bug and #839314.  If not,
> I will request that cairocffi drops its Build-Dependency and then remove
> xcffib plus haskell-xcb-types from testing.

I won't object to removal if it becomes necessary.

Thanks,
Iain.



signature.asc
Description: PGP signature

Bug#860739: libghc-xcb-types-dev: cannot parse xcb-proto 1.12

[Niels Thykier]

On Sat, 22 Apr 2017 18:13:50 +0100 James Cowgill 
wrote:
> Hi,
> 
> On Wed, 19 Apr 2017 16:05:12 +0100 James Cowgill  wrote:
> > Package: libghc-xcb-types-dev
> > Version: 0.7.1-4
> > Severity: grave
> > Tags: sid stretch
> > Control: block 839314 by -1
> > 
> > Hi,
> > 
> > Currently haskell-xcb-types cannot parse xcb-proto 1.12. While strictly
> > speaking you don't have to use this package to just parse the files from
> > xcb-proto, it is the package's main purpose and (afaik) only use in the
> > archive.
> 
> [...]
> 
> Relevant xcb-types commits:
> https://github.com/aslatter/xcb-types/commit/0991f1d61b92371e9af51ab0fa3699d7c32e2b65
> https://github.com/aslatter/xcb-types/commit/a86e578f0860670f3d43fa2d414e93a60aa72e2d
> 
> Commit which would need backporting to xcffib:
> https://github.com/tych0/xcffib/commit/6e6646b7d5eec253225579e1cbb40acf380648b4
> 
> (xcffib would then need to depend on xcb-types >= 0.8)
> 
> I notice the xcffib patches ignore the new alignment fields so we might
> be able to bodge it by ignoring the "required_start_align" elements in
> xcb-types, but since that's a bit of a hack, there are no other rdeps
> of xcb-types, and we're going to have to patch both packages anyway, I
> don't know if it's worth it.
> 
> James
> 

Hi James,

Thanks for the update and for looking into this issue.  I am glad to see
there is progress on this issue.

I hope we can get a timely resolution on this bug and #839314.  If not,
I will request that cairocffi drops its Build-Dependency and then remove
xcffib plus haskell-xcb-types from testing.

Thanks,
~Niels

Bug#860739: libghc-xcb-types-dev: cannot parse xcb-proto 1.12

[James Cowgill]

Hi,

On Wed, 19 Apr 2017 16:05:12 +0100 James Cowgill  wrote:
> Package: libghc-xcb-types-dev
> Version: 0.7.1-4
> Severity: grave
> Tags: sid stretch
> Control: block 839314 by -1
> 
> Hi,
> 
> Currently haskell-xcb-types cannot parse xcb-proto 1.12. While strictly
> speaking you don't have to use this package to just parse the files from
> xcb-proto, it is the package's main purpose and (afaik) only use in the
> archive.

On closer inspection, it looks like haskell-xcb-types has never been
able to parse the full xcb-proto. It's only ever been able to parse the
bits that xcffib needs to work properly. Since adding the extra support
is a lot of work and no packages will use it, I suggest that only the
xcffib stuff is fixed and then this bug can be cloned as important to
cover the other bits.

[...]
> This is the cause of the xcffib FTBFS in #839314
> 
> Apparently xcb-types 0.8.0 fixes this. I don't know how suitable that is
> to be included into stretch.

Upgrading to 0.8.0 breaks the API by adding alignment fields everywhere
to handle the parsing of the new "required_start_align" element. This
means we would also need to backport the relevant xcffib patches as
well. I think this is the best option though.

Relevant xcb-types commits:
https://github.com/aslatter/xcb-types/commit/0991f1d61b92371e9af51ab0fa3699d7c32e2b65
https://github.com/aslatter/xcb-types/commit/a86e578f0860670f3d43fa2d414e93a60aa72e2d

Commit which would need backporting to xcffib:
https://github.com/tych0/xcffib/commit/6e6646b7d5eec253225579e1cbb40acf380648b4

(xcffib would then need to depend on xcb-types >= 0.8)

I notice the xcffib patches ignore the new alignment fields so we might
be able to bodge it by ignoring the "required_start_align" elements in
xcb-types, but since that's a bit of a hack, there are no other rdeps
of xcb-types, and we're going to have to patch both packages anyway, I
don't know if it's worth it.

James



signature.asc
Description: OpenPGP digital signature

Bug#860739: libghc-xcb-types-dev: cannot parse xcb-proto 1.12

[James Cowgill]

Package: libghc-xcb-types-dev
Version: 0.7.1-4
Severity: grave
Tags: sid stretch
Control: block 839314 by -1

Hi,

Currently haskell-xcb-types cannot parse xcb-proto 1.12. While strictly
speaking you don't have to use this package to just parse the files from
xcb-proto, it is the package's main purpose and (afaik) only use in the
archive.

To reproduce (in ghci):
> import System.FilePath.Glob
> import Data.XCB.FromXML
>
> namesMatching "/usr/share/xcb/*.xml" >>= fromFiles

Gives:
> *** Exception: Invalid bitCase: QName {qName = "required_start_align", qURI = 
> Nothing, qPrefix = Nothing}
> CallStack (from HasCallStack):
>   error, called at ./Data/XCB/FromXML.hs:382:19 in 
> xcb-types-0.7.1-CupcnVsMv4F6MMa4qRj9I6:Data.XCB.FromXML

This is the cause of the xcffib FTBFS in #839314

Apparently xcb-types 0.8.0 fixes this. I don't know how suitable that is
to be included into stretch.

Thanks,
James



signature.asc
Description: OpenPGP digital signature