branch master updated: * tp/Texinfo/Convert/Unicode.pm (check_unicode_point_conversion): with strict encoding, there is a warning for nonchar with perl 5.10.1 on solaris 11, though not on solaris 10. So for a range of perl versions, 5.10.0 to 5.13.8 we do not attempt the eval and check for the codepoints to be below 0x10FFFF early.

[Patrice Dumas]

This is an automated email from the git hooks/post-receive script.

pertusus pushed a commit to branch master
in repository texinfo.

The following commit(s) were added to refs/heads/master by this push:
     new 30316397b4 * tp/Texinfo/Convert/Unicode.pm 
(check_unicode_point_conversion): with strict encoding, there is a warning for 
nonchar with perl 5.10.1 on solaris 11, though not on solaris 10.  So for a 
range of perl versions, 5.10.0 to 5.13.8 we do not attempt the eval and check 
for the codepoints to be below 0x10FFFF early.
30316397b4 is described below

commit 30316397b48707ad83b19a583dbc7ecbb94b164d
Author: Patrice Dumas <pertusus@free.fr>
AuthorDate: Sat Jul 29 09:51:35 2023 +0200

    * tp/Texinfo/Convert/Unicode.pm (check_unicode_point_conversion): with
    strict encoding, there is a warning for nonchar with perl 5.10.1 on
    solaris 11, though not on solaris 10.  So for a range of perl versions,
    5.10.0 to 5.13.8 we do not attempt the eval and check for the
    codepoints to be below 0x10FFFF early.
---
 ChangeLog                     |  8 ++++++++
 tp/Texinfo/Convert/Unicode.pm | 26 +++++++++++++++++++++++++-
 2 files changed, 33 insertions(+), 1 deletion(-)

diff --git a/ChangeLog b/ChangeLog
index 6ab14bb02b..26715d421c 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,11 @@
+2023-07-29  Patrice Dumas  <pertusus@free.fr>
+
+       * tp/Texinfo/Convert/Unicode.pm (check_unicode_point_conversion): with
+       strict encoding, there is a warning for nonchar with perl 5.10.1 on
+       solaris 11, though not on solaris 10.  So for a range of perl versions,
+       5.10.0 to 5.13.8 we do not attempt the eval and check for the
+       codepoints to be below 0x10FFFF early.
+
 2023-07-28  Gavin Smith <gavinsmith0123@gmail.com>
 
        * doc/texinfo.tex (\deftypeline): Read arguments correctly, in
diff --git a/tp/Texinfo/Convert/Unicode.pm b/tp/Texinfo/Convert/Unicode.pm
index f3cd447664..83812ca79e 100644
--- a/tp/Texinfo/Convert/Unicode.pm
+++ b/tp/Texinfo/Convert/Unicode.pm
@@ -1597,8 +1597,32 @@ sub check_unicode_point_conversion($;$)
   # The warning about non-characters is only given when the code
   # point is attempted to be output, not just manipulated.
   # 
http://stackoverflow.com/questions/5127725/how-could-i-catch-an-unicode-non-character-warning
+  # In perl 5.36.0, Encoding and printing also do not give a warning, so
+  # no warning for:
   #
-  # Therefore, we have to try to output it within an eval.
+  # my ($fh, $string);
+  # open($fh, ">", \$string);
+  # my $char = chr(hex("110000"));
+  # print $fh Encode::encode("utf-8", $char);
+  #
+  # but there is a warning if going through an encoding layer as below.
+  #
+  # In perl 5.10.1 on solaris 11, but not on solaris 10, the warning all does
+  # not catch the nonchar warning (this warning seems to be defined on
+  # 5.13.10 or newer).  This may be a consequence of what is described in
+  # 
http://stackoverflow.com/questions/5127725/how-could-i-catch-an-unicode-non-character-warning
+  # as a compiler bug, but it is unclear.  This does not happen with the
+  # lax conversion to utf8, but we prefer to use a strict conversion.
+  #
+  # To avoid outputting a warning, we do not even try the eval for perls
+  # in the 5.10.0 5.13.8 range
+  if ($] >= 5.010 and $] <= 5.013008) {
+    if (hex($arg) > 0x10FFFF) {
+      return 0;
+    }
+  }
+  #
+  # For the other cases, we have to try to output it within an eval.
   # Since opening /dev/null or a temporary file means
   # more system-dependent checks, use a string as our
   # filehandle.