mbiter, mbuiter: Improve state handling after invalid input

[Bruno Haible]

After mbrtowc or mbrtoc32 failed with return value (size_t) -1, we don't know in
which state the mbstate_t is. Therefore it's best to clear it before potentially
calling mbrtowc or mbrtoc32 again.


2023-07-04  Bruno Haible  <bruno@clisp.org>

        mbiter, mbuiter, mbfile: Improve state handling after invalid input.
        * lib/mbiter.h (mbiter_multi_next): After an invalid multibyte sequence
        was encountered, clear the state.
        * lib/mbuiter.h (mbuiter_multi_next): Likewise.
        * lib/mbfile.h (mbfile_multi_getc): Likewise.

diff --git a/lib/mbfile.h b/lib/mbfile.h
index 9a2532992e..7670dabfbb 100644
--- a/lib/mbfile.h
+++ b/lib/mbfile.h
@@ -147,6 +147,8 @@ mbfile_multi_getc (struct mbchar *mbc, struct mbfile_multi 
*mbf)
           /* Return a single byte.  */
           bytes = 1;
           mbc->wc_valid = false;
+          /* Allow the next invocation to continue from a sane state.  */
+          memset (&mbf->state, '\0', sizeof (mbstate_t));
           break;
         }
       else if (bytes == (size_t) -2)
diff --git a/lib/mbiter.h b/lib/mbiter.h
index bc88b4f3a2..e4963010f3 100644
--- a/lib/mbiter.h
+++ b/lib/mbiter.h
@@ -146,8 +146,9 @@ mbiter_multi_next (struct mbiter_multi *iter)
           /* An invalid multibyte sequence was encountered.  */
           iter->cur.bytes = 1;
           iter->cur.wc_valid = false;
-          /* Whether to set iter->in_shift = false and reset iter->state
-             or not is not very important; the string is bogus anyway.  */
+          /* Allow the next invocation to continue from a sane state.  */
+          iter->in_shift = false;
+          memset (&iter->state, '\0', sizeof (mbstate_t));
         }
       else if (iter->cur.bytes == (size_t) -2)
         {
diff --git a/lib/mbuiter.h b/lib/mbuiter.h
index 93dec81603..e0fcd03d3d 100644
--- a/lib/mbuiter.h
+++ b/lib/mbuiter.h
@@ -155,8 +155,9 @@ mbuiter_multi_next (struct mbuiter_multi *iter)
           /* An invalid multibyte sequence was encountered.  */
           iter->cur.bytes = 1;
           iter->cur.wc_valid = false;
-          /* Whether to set iter->in_shift = false and reset iter->state
-             or not is not very important; the string is bogus anyway.  */
+          /* Allow the next invocation to continue from a sane state.  */
+          iter->in_shift = false;
+          memset (&iter->state, '\0', sizeof (mbstate_t));
         }
       else if (iter->cur.bytes == (size_t) -2)
         {