Problems building GC 18.1 on CentOS 7

Problems building GC 18.1 on CentOS 7

[Tom Kacvinsky via Gdb]

Hi,

I found a few issues with building GDB 18.1 on CentOS 7, using
GCC 16.1.0 (which I had built from source)  They are two fold

1. An issue with ./gdb/tui/tui-io.c using macro define(s) that the
system curses headers don't provide.

2. An issue with ptrace macro define(s) that are provided in glibc
2.18 or later, but not on CentOS 7, which is glibc 2.17 based.

I used Claude Opus 5 to generate a patch and manually inspected it.
I am not sure what the GDB policy is for AI assisted patches.

I will get the exact compilation failures and the patch to the list on
Monday.  To be honest, when it comes to the patch, I think I need to
be mindful that I am building on Linux (via __linux__), using GCC
(via __GNU__) and the glibc version is less than 2.18 using
__GLIBC_PREREQ from features.h

Regards,

Tom

Re: Problems building GC 18.1 on CentOS 7

[Tom Kacvinsky via Gdb]

On Sat, Oct 3, 2026 at 3:18 PM Tom Kacvinsky <tkacvins@gmail.com> wrote:

<snip>

Corrections

(a) The subject should have used GDB 18.1, not GC 18.1

(b) I meant __GNUC__, not __GNU__

(c) Perhaps checks would be better done through autoconf.  For instance,
using autoconf to detect if the ptrace specific macros are available, etc...

I'll examine option (c) when I am back in the office.

Tom

Re: Problems building GC 18.1 on CentOS 7

[Simon Marchi via Gdb]

On 2026-10-03 15:18, Tom Kacvinsky via Gdb wrote:
> Hi,
> 
> I found a few issues with building GDB 18.1 on CentOS 7, using
> GCC 16.1.0 (which I had built from source)  They are two fold
> 
> 1. An issue with ./gdb/tui/tui-io.c using macro define(s) that the
> system curses headers don't provide.
> 
> 2. An issue with ptrace macro define(s) that are provided in glibc
> 2.18 or later, but not on CentOS 7, which is glibc 2.17 based.
> 
> I used Claude Opus 5 to generate a patch and manually inspected it.
> I am not sure what the GDB policy is for AI assisted patches.
> 
> I will get the exact compilation failures and the patch to the list on
> Monday.  To be honest, when it comes to the patch, I think I need to
> be mindful that I am building on Linux (via __linux__), using GCC
> (via __GNU__) and the glibc version is less than 2.18 using
> __GLIBC_PREREQ from features.h
> 
> Regards,
> 
> Tom

Hi Tom,

Could you please open a bug on Bugzilla?

https://sourceware.org/bugzilla/

If we are going to make a fix targetting 18.2, we will need a bug number
attached in any case.

Thanks,

Simon

Re: Problems building GC 18.1 on CentOS 7

[Tom Kacvinsky via Gdb]

Hi Simon,

> On Oct 3, 2026, at 18:05, Simon Marchi <simark@simark.ca> wrote:
> 
>> Could you please open a bug on Bugzilla?
> 
> https://sourceware.org/bugzilla/
> 
> If we are going to make a fix targetting 18.2, we will need a bug number attached in any case.

Will do!

Tom

Re: Problems building GC 18.1 on CentOS 7

[Tom Kacvinsky via Gdb]

Hi Simon,

On Sat, Oct 3, 2026 at 6:25 PM Tom Kacvinsky <tkacvins@gmail.com> wrote:

>
> Hi Simon,
>
> > On Oct 3, 2026, at 18:05, Simon Marchi <simark@simark.ca> wrote:
> >
> >> Could you please open a bug on Bugzilla?
> >
> > https://sourceware.org/bugzilla/
> >
> > If we are going to make a fix targetting 18.2, we will need a bug number
> attached in any case.
>
> Will do!
>

I have filed

https://sourceware.org/bugzilla/show_bug.cgi?id=34708

with proposed ML generated patches.

Note that those patches apply against HEAD, and that I did not regenerate
the configure
script as part of the patches.

I have a strong suspicion that the change made to configure.ac for the
curses issue could
be improved.

Re: Problems building GC 18.1 on CentOS 7

[Guinevere Larsen via Gdb]

On 10/3/26 4:18 PM, Tom Kacvinsky via Gdb wrote:
> I used Claude Opus 5 to generate a patch and manually inspected it.
> I am not sure what the GDB policy is for AI assisted patches.
As I understand it, we're not allowing any AI code in the project, as we 
don't know to whom the copyright belongs, and thus can't be sure that it 
can be transferred to the FSF

-- 
Cheers,
Guinevere Larsen
it/its
she/her (deprecated)

>
> I will get the exact compilation failures and the patch to the list on
> Monday.  To be honest, when it comes to the patch, I think I need to
> be mindful that I am building on Linux (via __linux__), using GCC
> (via __GNU__) and the glibc version is less than 2.18 using
> __GLIBC_PREREQ from features.h
>
> Regards,
>
> Tom

Re: Problems building GC 18.1 on CentOS 7

[Tom Kacvinsky via Gdb]

Hi Guinevere,

On Mon, Oct 5, 2026 at 7:28 AM Guinevere Larsen <guinevere@redhat.com> wrote:
>
>
> On 10/3/26 4:18 PM, Tom Kacvinsky via Gdb wrote:
> > I used Claude Opus 5 to generate a patch and manually inspected it.
> > I am not sure what the GDB policy is for AI assisted patches.
> As I understand it, we're not allowing any AI code in the project, as we
> don't know to whom the copyright belongs, and thus can't be sure that it
> can be transferred to the FSF
>

I think that is completely fair.  If you go look at the bug I filed,
Simon cannot
reproduce the problem, and I think that is because I am behind on the CentOS
version I am using.  So I think the best bet is just to close the bug
with "won't
fix" and then I'll just carry on with my patches internal to my employer.  We do
not ship gdb - so we are not in violation of the GPL license. I am just building
the latest, greatest for use by developers.

Thanks,

Tom