[PATCH 1/1] gdb/tui: Fix build for older ncurses

[PATCH 1/1] gdb/tui: Fix build for older ncurses

[Ciaran Woodward]

Older versions of ncurses (including the version that ships inside
macos, and Centos 7) do not include the A_ITALIC macro. This patch
simply hides any use of A_ITALIC behind a preprocessor guard.

The result of this is that italics won't be rendered in the tui
if ncurses isn't supported. We do have other options if we think
it's important - for instance we could show italics as bold if
italics aren't supported. From my understanding, that might be
overthinking it - so I took the simplest approach here, just to
fix the build.

Those versions also define tgetnum as:
  int tgetnum(char *id);
so attempting to compile for c++ results in the error:
  ISO C++ forbids converting a string constant to 'char*' [-Werror=write-strings]

This is just a dated API issue, so a const cast resolves the issue.
---
 gdb/tui/tui-io.c | 4 ++++
 gdb/ui-style.c   | 2 +-
 2 files changed, 5 insertions(+), 1 deletion(-)

diff --git a/gdb/tui/tui-io.c b/gdb/tui/tui-io.c
index c97e8fd1717..84cad9366d5 100644
--- a/gdb/tui/tui-io.c
+++ b/gdb/tui/tui-io.c
@@ -319,7 +319,9 @@ tui_apply_style (WINDOW *w, ui_file_style style)
   wattron (w, A_NORMAL);
   wattroff (w, A_BOLD);
   wattroff (w, A_DIM);
+#ifdef  A_ITALIC
   wattroff (w, A_ITALIC);
+#endif
   wattroff (w, A_UNDERLINE);
   wattroff (w, A_REVERSE);
   if (last_color_pair != -1)
@@ -368,8 +370,10 @@ tui_apply_style (WINDOW *w, ui_file_style style)
       gdb_assert_not_reached ("invalid intensity");
     }
 
+#ifdef  A_ITALIC
   if (style.is_italic ())
     wattron (w, A_ITALIC);
+#endif
 
   if (style.is_underline ())
     wattron (w, A_UNDERLINE);
diff --git a/gdb/ui-style.c b/gdb/ui-style.c
index 9a58e4dd2ae..ccc9094ca59 100644
--- a/gdb/ui-style.c
+++ b/gdb/ui-style.c
@@ -594,7 +594,7 @@ colorsupport ()
     {
       std::vector<color_space> result = {color_space::MONOCHROME};
 
-      int colors = tgetnum ("Co");
+      int colors = tgetnum (const_cast<char*> ("Co"));
       if (colors >= 8)
 	result.push_back (color_space::ANSI_8COLOR);
       if (colors >= 16)
-- 
2.25.1

Re: [PATCH 1/1] gdb/tui: Fix build for older ncurses

[Tom Tromey]

>>>>> "Ciaran" == Ciaran Woodward <ciaranwoodward@xmos.com> writes:

Ciaran> Older versions of ncurses (including the version that ships inside
Ciaran> macos, and Centos 7) do not include the A_ITALIC macro. This patch
Ciaran> simply hides any use of A_ITALIC behind a preprocessor guard.

Thanks for doing this.

Ciaran> The result of this is that italics won't be rendered in the tui
Ciaran> if ncurses isn't supported. We do have other options if we think
Ciaran> it's important - for instance we could show italics as bold if
Ciaran> italics aren't supported. From my understanding, that might be
Ciaran> overthinking it - so I took the simplest approach here, just to
Ciaran> fix the build.

Seems reasonable enough.

Ciaran> Those versions also define tgetnum as:
Ciaran>   int tgetnum(char *id);
Ciaran> so attempting to compile for c++ results in the error:
Ciaran>   ISO C++ forbids converting a string constant to 'char*' [-Werror=write-strings]

Ciaran> diff --git a/gdb/ui-style.c b/gdb/ui-style.c
Ciaran> index 9a58e4dd2ae..ccc9094ca59 100644
Ciaran> --- a/gdb/ui-style.c
Ciaran> +++ b/gdb/ui-style.c
Ciaran> @@ -594,7 +594,7 @@ colorsupport ()
Ciaran>      {
Ciaran>        std::vector<color_space> result = {color_space::MONOCHROME};
 
Ciaran> -      int colors = tgetnum ("Co");
Ciaran> +      int colors = tgetnum (const_cast<char*> ("Co"));

Would you mind adding a comment here explaining the need for the cast
and also for which version of curses it was found to be needed?
I think that may help future developers.

Anyway ok with this change.
Approved-By: Tom Tromey <tom@tromey.com>

Tom

RE: [PATCH 1/1] gdb/tui: Fix build for older ncurses

[Ciaran Woodward]

> Would you mind adding a comment here explaining the need for the cast
> and also for which version of curses it was found to be needed?
> I think that may help future developers.
> 
> Anyway ok with this change.
> Approved-By: Tom Tromey <tom@tromey.com>

Thanks - added the comment & pushed.

Re: [PATCH 1/1] gdb/tui: Fix build for older ncurses

[Daniel Fellows]

Hi Ciaran,

Would you be willing to backport this fix to gdb-17-branch?

We encountered the A_ITALIC build failure after moving our macOS builds
to GDB 17. I tested the fix successfully on both Intel and Arm macOS
hosts.

There is a small conflict with commit 07e542a3bd05, which added MinGW
handling in the same section of ui-style.c. The resolution is to retain
that MinGW block alongside your tgetnum change.

Thanks,
Daniel

RE: [PATCH 1/1] gdb/tui: Fix build for older ncurses

[Ciaran Woodward]

> -----Original Message-----
> From: Daniel Fellows <daniel.fellows@arm.com>
> Sent: 27 July 2026 17:16
>  
> Would you be willing to backport this fix to gdb-17-branch?
> 
> We encountered the A_ITALIC build failure after moving our macOS builds
> to GDB 17. I tested the fix successfully on both Intel and Arm macOS
> hosts.
> 
> There is a small conflict with commit 07e542a3bd05, which added MinGW
> handling in the same section of ui-style.c. The resolution is to retain
> that MinGW block alongside your tgetnum change.

Hi Daniel,

I'm happy for this fix to be backported, and I'm happy to push the commit
if you aren't able - from my perspective it seems extremely low risk.

I'm not quite sure on the backport process - I assume it needs approval by
a maintainer, at least.

Can any maintainer fill us in?

Cheers,
Ciaran

Re: [PATCH 1/1] gdb/tui: Fix build for older ncurses

[Daniel Fellows]

Hi Ciaran,

Thanks for getting back to me. Yes, if you could push the backport once
the required approval is confirmed, that would be great.

Hopefully one of the maintainers can confirm the correct backport
process and provide approval.

Cheers,
Daniel