[nasm:nasm-2.15.xx] preproc: handle %+ pasting after empty expansions

[nasm-bot for H. Peter Anvin (Intel)]

Commit-ID:  122c5fb75986adc37dfb147cc2a613e3ebc66e80
Gitweb:     http://repo.or.cz/w/nasm.git?a=commitdiff;h=122c5fb75986adc37dfb147cc2a613e3ebc66e80
Author:     H. Peter Anvin (Intel) <hpa at zytor.com>
AuthorDate: Sun, 5 Jul 2020 03:39:04 -0700
Committer:  H. Peter Anvin (Intel) <hpa at zytor.com>
CommitDate: Sun, 5 Jul 2020 03:39:04 -0700

preproc: handle %+ pasting after empty expansions

%+ tokens can end up next to each other, or at the beginning or the
end of an expansion if we try to paste the output of empty
macros. This is perhaps particularly likely to happen in %[]
expressions.

Signed-off-by: H. Peter Anvin (Intel) <hpa at zytor.com>


---
 asm/preproc.c | 36 ++++++++++++++++--------------------
 1 file changed, 16 insertions(+), 20 deletions(-)

diff --git a/asm/preproc.c b/asm/preproc.c
index 5cb92879..81c72042 100644
--- a/asm/preproc.c
+++ b/asm/preproc.c
@@ -4726,9 +4726,8 @@ static bool paste_tokens(Token **head, const struct tokseq_match *m,
             if (!handle_explicit)
                 break;
 
-            /* Left pasting token is start of line */
+            /* Left pasting token is start of line, just drop %+ */
             if (!prev_nonspace) {
-                nasm_nonfatal("No lvalue found on pasting");
                 tok = delete_Token(tok);
                 break;
             }
@@ -4741,28 +4740,25 @@ static bool paste_tokens(Token **head, const struct tokseq_match *m,
             /* Delete leading whitespace */
             next = zap_white(t->next);
 
-            /* Delete the %+ token itself */
-            nasm_assert(next == tok);
-            next = delete_Token(next);
-
-            /* Delete trailing whitespace */
-            next = zap_white(next);
+            /*
+             * Delete the %+ token itself, followed by any whitespace.
+             * In a sequence of %+ ... %+ ... %+ pasting sequences where
+             * some expansions in the middle have ended up empty,
+             * we can end up having multiple %+ tokens in a row;
+             * just drop whem in that case.
+             */
+            while (next) {
+                if (next->type == TOK_PASTE || next->type == TOK_WHITESPACE)
+                    next = delete_Token(next);
+                else
+                    break;
+            }
 
             /*
-             * No ending token, this might happen in two
-             * cases
-             *
-             *  1) There indeed no right token at all
-             *  2) There is a bare "%define ID" statement,
-             *     and @ID does expand to whitespace.
-             *
-             * So technically we need to do a grammar analysis
-             * in another stage of parsing, but for now lets don't
-             * change the behaviour people used to. Simply allow
-             * whitespace after paste token.
+             * Nothing after? Just leave the existing token.
              */
             if (!next) {
-                *prev_nonspace = tok = NULL; /* End of line */
+                t->next = tok = NULL; /* End of line */
                 break;
             }