Attempting to fix hyperscan FTBFS

[jt]

Hi All,

I am looking at the FTBFS for hyperscan[0] and need some assistance.
From what I can tell it looks like this may be related to the --as-
needed linker change.

The reason I say it may be related to the linking change is I can build
the 5.1.0 release against the fedora-29-x86_64 mock profile and the
CMakeOutput.log from a local mock build against the fedora-rawhide-
x86_64 profile[1] seems to indicate a linking issue.

This is outside my expertise but I would obviously like to fix the
issue the most appropriate way possible. 

So if this is linker issue, what's the best approach to fix in this
case?

If it isn't a linker issue, where should I start poking next?

Thanks in advance for any pointers.

[0] https://koji.fedoraproject.org/koji/taskinfo?taskID=32400415
[1] https://gist.github.com/jmtaylor90/bd9680de68d1f8486dacd4e0b645f07b


JT

[Gwyn Ciesla]

‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐
On Tuesday, February 12, 2019 2:42 PM, jt <jtfas90(a)gmail.com&gt; wrote:

>  Hi All,
>   
>  I am looking at the FTBFS for hyperscan[0] and need some assistance.
>  From what I can tell it looks like this may be related to the --as-
>  needed linker change.
>   
>  The reason I say it may be related to the linking change is I can
> build
>  the 5.1.0 release against the fedora-29-x86_64 mock profile and the
>  CMakeOutput.log from a local mock build against the fedora-rawhide-
>  x86_64 profile[1] seems to indicate a linking issue.
>   
>  This is outside my expertise but I would obviously like to fix the
>  issue the most appropriate way possible.
>   
>  So if this is linker issue, what's the best approach to fix in
> this
>  case?
>   
>  If it isn't a linker issue, where should I start poking next?
>   
>  Thanks in advance for any pointers.
>   
>  [0] https://koji.fedoraproject.org/koji/taskinfo?taskID=32400415
>  [1] https://gist.github.com/jmtaylor90/bd9680de68d1f8486dacd4e0b645f07b
>   
>  JT
>   
>  devel mailing list -- devel(a)lists.fedoraproject.org
>  To unsubscribe send an email to devel-leave(a)lists.fedoraproject.org
>  Fedora Code of Conduct: https://getfedora.org/code-of-conduct.html
>  List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines
>  List Archives:
> https://lists.fedoraproject.org/archives/list/[e-mail] 
I've hit this as well on other pacakges.

[Tom Hughes]

On 12/02/2019 20:42, jt wrote:

>  I am looking at the FTBFS for hyperscan[0] and need some assistance.
>   From what I can tell it looks like this may be related to the --as-
>  needed linker change. 
Why do you think that? The koji log shows that it is failing
because it is asking for pcre 8.41 and won't accept any other
version, older or newer. As we have 8.43 it is failing.

I'd try patching it to >=8.41 instead of =8.41 and it will
most likely be fine - then suggest to upstream that they not
be quite so daft as to require an exact version...

Tom

-- 
Tom Hughes (tom(a)compton.nu)
http://compton.nu/

[Pete Walter]

Looking at the build log, it seems to be checking for 'libpcre = 8.41', but the
build root has 8.43-RC1 which causes the build to error out.

Pete

12.02.2019, 20:43, "jt" <jtfas90(a)gmail.com&gt;:
> Hi All,
>
> I am looking at the FTBFS for hyperscan[0] and need some assistance.
> From what I can tell it looks like this may be related to the --as-
> needed linker change.
>
> The reason I say it may be related to the linking change is I can build
> the 5.1.0 release against the fedora-29-x86_64 mock profile and the
> CMakeOutput.log from a local mock build against the fedora-rawhide-
> x86_64 profile[1] seems to indicate a linking issue.
>
> This is outside my expertise but I would obviously like to fix the
> issue the most appropriate way possible.
>
> So if this is linker issue, what's the best approach to fix in this
> case?
>
> If it isn't a linker issue, where should I start poking next?
>
> Thanks in advance for any pointers.
>
> [0] https://koji.fedoraproject.org/koji/taskinfo?taskID=32400415
> [1] https://gist.github.com/jmtaylor90/bd9680de68d1f8486dacd4e0b645f07b
>
> JT
>
> _______________________________________________
> devel mailing list -- devel(a)lists.fedoraproject.org
> To unsubscribe send an email to devel-leave(a)lists.fedoraproject.org
> Fedora Code of Conduct: https://getfedora.org/code-of-conduct.html
> List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines
> List Archives:
https://lists.fedoraproject.org/archives/list/[e-mail]

[jt]

On Tue, 2019-02-12 at 21:06 +0000, Pete Walter wrote:
>  Looking at the build log, it seems to be checking for 'libpcre =
>  8.41', but the build root has 8.43-RC1 which causes the build to
>  error out.
>  
>  Pete
>  
>  12.02.2019, 20:43, "jt" <jtfas90(a)gmail.com&gt;:
>  > Hi All,
>  > 
>  > I am looking at the FTBFS for hyperscan[0] and need some
>  > assistance.
>  > From what I can tell it looks like this may be related to the --as-
>  > needed linker change.
>  > 
>  > The reason I say it may be related to the linking change is I can
>  > build
>  > the 5.1.0 release against the fedora-29-x86_64 mock profile and the
>  > CMakeOutput.log from a local mock build against the fedora-rawhide-
>  > x86_64 profile[1] seems to indicate a linking issue.
>  > 
>  > This is outside my expertise but I would obviously like to fix the
>  > issue the most appropriate way possible.
>  > 
>  > So if this is linker issue, what's the best approach to fix in this
>  > case?
>  > 
>  > If it isn't a linker issue, where should I start poking next?
>  > 
>  > Thanks in advance for any pointers.
>  > 
>  > [0] https://koji.fedoraproject.org/koji/taskinfo?taskID=32400415
>  > [1] 
>  > https://gist.github.com/jmtaylor90/bd9680de68d1f8486dacd4e0b645f07b
>  > 
>  > JT
>  > 
>  >  
My rereads before sending failed me.

I linked the original 5.0.0 build, this is the proper 5.1.0 build link:

https://koji.fedoraproject.org/koji/taskinfo?=32766510

The 5.1.0 release fixes the pcre explicit version requirement among
other things.

Sorry for the confusion

JT

[jt]

On Tue, 2019-02-12 at 16:13 -0500, jt wrote:
>  On Tue, 2019-02-12 at 21:06 +0000, Pete Walter wrote:
>  > Looking at the build log, it seems to be checking for 'libpcre =
>  > 8.41', but the build root has 8.43-RC1 which causes the build to
>  > error out.
>  > 
>  > Pete
>  > 
>  > 12.02.2019, 20:43, "jt" <jtfas90(a)gmail.com&gt;:
>  > > Hi All,
>  > > 
>  > > I am looking at the FTBFS for hyperscan[0] and need some
>  > > assistance.
>  > > From what I can tell it looks like this may be related to the --
>  > > as-
>  > > needed linker change.
>  > > 
>  > > The reason I say it may be related to the linking change is I can
>  > > build
>  > > the 5.1.0 release against the fedora-29-x86_64 mock profile and
>  > > the
>  > > CMakeOutput.log from a local mock build against the fedora-
>  > > rawhide-
>  > > x86_64 profile[1] seems to indicate a linking issue.
>  > > 
>  > > This is outside my expertise but I would obviously like to fix
>  > > the
>  > > issue the most appropriate way possible.
>  > > 
>  > > So if this is linker issue, what's the best approach to fix in
>  > > this
>  > > case?
>  > > 
>  > > If it isn't a linker issue, where should I start poking next?
>  > > 
>  > > Thanks in advance for any pointers.
>  > > 
>  > > [0] https://koji.fedoraproject.org/koji/taskinfo?taskID=32400415
>  > > [1] 
>  > > https://gist.github.com/jmtaylor90/bd9680de68d1f8486dacd4e0b645f07b
>  > > 
>  > > JT
>  > > 
>  > > 
>  
>  My rereads before sending failed me.
>  
>  I linked the original 5.0.0 build, this is the proper 5.1.0 build
>  link:
>  
>  https://koji.fedoraproject.org/koji/taskinfo?=32766510
>  
>  The 5.1.0 release fixes the pcre explicit version requirement among
>  other things.
>  
>  Sorry for the confusion 
Sigh, that should be:

https://koji.fedoraproject.org/koji/taskinfo?taskID=32766510

JT

[Pete Walter]

12.02.2019, 21:16, "jt" <jtfas90(a)gmail.com&gt;:
>  On Tue, 2019-02-12 at 16:13 -0500, jt wrote:
> >  On Tue, 2019-02-12 at 21:06 +0000, Pete Walter wrote:
> >  > Looking at the build log, it seems to be checking for 'libpcre =
> >  > 8.41', but the build root has 8.43-RC1 which causes the build to
> >  > error out.
> >  >
> >  > Pete
> >  >
> >  > 12.02.2019, 20:43, "jt" <jtfas90(a)gmail.com&gt;:
> >  > > Hi All,
> >  > >
> >  > > I am looking at the FTBFS for hyperscan[0] and need some
> >  > > assistance.
> >  > > From what I can tell it looks like this may be related to the --
> >  > > as-
> >  > > needed linker change.
> >  > >
> >  > > The reason I say it may be related to the linking change is I can
> >  > > build
> >  > > the 5.1.0 release against the fedora-29-x86_64 mock profile and
> >  > > the
> >  > > CMakeOutput.log from a local mock build against the fedora-
> >  > > rawhide-
> >  > > x86_64 profile[1] seems to indicate a linking issue.
> >  > >
> >  > > This is outside my expertise but I would obviously like to fix
> >  > > the
> >  > > issue the most appropriate way possible.
> >  > >
> >  > > So if this is linker issue, what's the best approach to fix in
> >  > > this
> >  > > case?
> >  > >
> >  > > If it isn't a linker issue, where should I start poking next?
> >  > >
> >  > > Thanks in advance for any pointers.
> >  > >
> >  > > [0] https://koji.fedoraproject.org/koji/taskinfo?taskID=32400415
> >  > > [1]
> >  > > https://gist.github.com/jmtaylor90/bd9680de68d1f8486dacd4e0b645f07b
> >  > >
> >  > > JT
> >  > >
> >  > >
> >
> >  My rereads before sending failed me.
> >
> >  I linked the original 5.0.0 build, this is the proper 5.1.0 build
> >  link:
> >
> >  https://koji.fedoraproject.org/koji/taskinfo?=32766510
> >
> >  The 5.1.0 release fixes the pcre explicit version requirement among
> >  other things.
> >
> >  Sorry for the confusion
>
>  Sigh, that should be:
>
>  https://koji.fedoraproject.org/koji/taskinfo?taskID=32766510 
This log has "-mtune=option: not valid". I guess something has changed here with
gcc 9 that the cmake script doesn't like?

Pete

[Pete Walter]

12.02.2019, 21:16, "jt" <jtfas90(a)gmail.com&gt;:
>  On Tue, 2019-02-12 at 16:13 -0500, jt wrote:
> >  On Tue, 2019-02-12 at 21:06 +0000, Pete Walter wrote:
> >  > Looking at the build log, it seems to be checking for 'libpcre =
> >  > 8.41', but the build root has 8.43-RC1 which causes the build to
> >  > error out.
> >  >
> >  > Pete
> >  >
> >  > 12.02.2019, 20:43, "jt" <jtfas90(a)gmail.com&gt;:
> >  > > Hi All,
> >  > >
> >  > > I am looking at the FTBFS for hyperscan[0] and need some
> >  > > assistance.
> >  > > From what I can tell it looks like this may be related to the --
> >  > > as-
> >  > > needed linker change.
> >  > >
> >  > > The reason I say it may be related to the linking change is I can
> >  > > build
> >  > > the 5.1.0 release against the fedora-29-x86_64 mock profile and
> >  > > the
> >  > > CMakeOutput.log from a local mock build against the fedora-
> >  > > rawhide-
> >  > > x86_64 profile[1] seems to indicate a linking issue.
> >  > >
> >  > > This is outside my expertise but I would obviously like to fix
> >  > > the
> >  > > issue the most appropriate way possible.
> >  > >
> >  > > So if this is linker issue, what's the best approach to fix in
> >  > > this
> >  > > case?
> >  > >
> >  > > If it isn't a linker issue, where should I start poking next?
> >  > >
> >  > > Thanks in advance for any pointers.
> >  > >
> >  > > [0] https://koji.fedoraproject.org/koji/taskinfo?taskID=32400415
> >  > > [1]
> >  > > https://gist.github.com/jmtaylor90/bd9680de68d1f8486dacd4e0b645f07b
> >  > >
> >  > > JT
> >  > >
> >  > >
> >
> >  My rereads before sending failed me.
> >
> >  I linked the original 5.0.0 build, this is the proper 5.1.0 build
> >  link:
> >
> >  https://koji.fedoraproject.org/koji/taskinfo?=32766510
> >
> >  The 5.1.0 release fixes the pcre explicit version requirement among
> >  other things.
> >
> >  Sorry for the confusion
>
>  Sigh, that should be:
>
>  https://koji.fedoraproject.org/koji/taskinfo?taskID=32766510 
This log has "-mtune=option: not valid". I guess something has changed here with
gcc 9 that the cmake script doesn't like?

Pete

[Björn 'besser82' Esser]

>  This log has "-mtune=option: not valid". I guess something
> has changed
>  here with gcc 9 that the cmake script doesn't like? 

I've tailored a patch for CMakeLists.txt to fix the build [1].  If you
agree, I'll push to SCM (including update to v5.1.0) and build it.

Björn


[1]  https://koji.fedoraproject.org/koji/taskinfo?taskID=32767716

[Jason Taylor]

On Tue, Feb 12, 2019, 16:35 Björn 'besser82' Esser <
besser82(a)fedoraproject.org&gt; wrote:

>  > This log has "-mtune=option: not valid". I guess
> something has changed
>  > here with gcc 9 that the cmake script doesn't like?
>
>
>  I've tailored a patch for CMakeLists.txt to fix the build [1].  If you
>  agree, I'll push to SCM (including update to v5.1.0) and build it.
>
>  Björn
>
>
>  [1]  https://koji.fedoraproject.org/koji/taskinfo?taskID=32767716 

That works for me, thank you!

JT

[Björn 'besser82' Esser]

Am Dienstag, den 12.02.2019, 16:40 -0500 schrieb Jason Taylor:
>  
>  
>  On Tue, Feb 12, 2019, 16:35 Björn 'besser82' Esser <
>  besser82(a)fedoraproject.org&gt; wrote:
>  > > This log has "-mtune=option: not valid". I guess something has
>  > changed
>  > > here with gcc 9 that the cmake script doesn't like?
>  > 
>  > 
>  > I've tailored a patch for CMakeLists.txt to fix the build [1].  If
>  > you
>  > agree, I'll push to SCM (including update to v5.1.0) and build it.
>  > 
>  > Björn
>  > 
>  > 
>  > [1]  https://koji.fedoraproject.org/koji/taskinfo?taskID=32767716
>  
>  That works for me, thank you!
>  
>  JT 

Allrighty!  Here you go:

https://koji.fedoraproject.org/koji/buildinfo?buildID=1208802