Hi. Thanks for the report.
Gianfranco Costamagna writes ("Bug#968734: chiark-tcl: FTBFS with optimization
level -O3 and gcc-10 on s390x"):
> Hello, In Ubuntu the package FTBFS on s390x, because of -O3, and some non
> initialized variables.
How odd.
> This makes no sense, because the variables are passed by reference
> to the functions, but gcc is probably not smart enough to detect it.
I think it is more likely that it can see into blockcipher_prep, but
not follow the control flow. Perhaps, for example, it doesn't know
that cht_staticerr always returns nonzero, and it thinks that those
error paths in blockcipher_prep might end up using the values in the
caller.
Instead of your suggestion, if you can easily do so, can you try
this ?
Regards,
Ian.
>From 020f536563e566c6a17eeb790d2a5e56141e2b74 Mon Sep 17 00:00:00 2001
From: Ian Jackson <ijack...@chiark.greenend.org.uk>
Date: Thu, 20 Aug 2020 19:18:14 +0100
Subject: [PATCH] blockcipher_prep: Initialise out parameters to placate gcc
These are all set on the success exit path but GCC is not clever
enough to see this.
Closes: #968734
Signed-off-by: Ian Jackson <ijack...@chiark.greenend.org.uk>
---
crypto/crypto.c | 8 ++++++++
1 file changed, 8 insertions(+)
diff --git a/crypto/crypto.c b/crypto/crypto.c
index aee2556..6efea24 100644
--- a/crypto/crypto.c
+++ b/crypto/crypto.c
@@ -252,6 +252,14 @@ static int blockcipher_prep(Tcl_Interp *ip, Tcl_Obj
*key_obj,
int rc;
CiphKeyValue *key;
+ /* placate gcc, see Debian #968734 */
+ *key_r= 0;
+ *sched_r= 0;
+ *iv_r= 0;
+ *iv_lenbytes_r= 0;
+ *buffers_r= 0;
+ *nblocks_r= 0;
+
if (data_len % alg->blocksize)
return cht_staticerr(ip, "block cipher input not whole number of blocks",
"HBYTES BLOCKCIPHER LENGTH");
--
2.20.1
