* tic4x-coff: ubsan: various shift UB
@ 2020-09-01 6:47 Alan Modra
0 siblings, 0 replies; 2+ messages in thread
From: Alan Modra @ 2020-09-01 6:47 UTC (permalink / raw)
To: binutils
* config/tc-tic4x.c (tic4x_gen_to_words): Rewrite mantissa
overflow test without UB. Avoid other UB shifts by making them
unsigned.
diff --git a/gas/config/tc-tic4x.c b/gas/config/tc-tic4x.c
index 1889b47128..4efb9d4ae5 100644
--- a/gas/config/tc-tic4x.c
+++ b/gas/config/tc-tic4x.c
@@ -453,8 +453,8 @@ tic4x_gen_to_words (FLONUM_TYPE flonum, LITTLENUM_TYPE *words, int precision)
/* Store the mantissa data into smant and the roundbit into rbit */
for (p = flonum.leader; p >= flonum.low && shift > -16; p--)
{
- tmp = shift >= 0 ? *p << shift : *p >> -shift;
- rbit = shift < 0 ? ((*p >> (-shift-1)) & 0x1) : 0;
+ tmp = shift >= 0 ? (unsigned) *p << shift : (unsigned) *p >> -shift;
+ rbit = shift < 0 ? (((unsigned) *p >> (-shift-1)) & 0x1) : 0;
smant |= tmp;
shift -= 16;
}
@@ -463,18 +463,14 @@ tic4x_gen_to_words (FLONUM_TYPE flonum, LITTLENUM_TYPE *words, int precision)
if(rbit)
{
/* If the mantissa is going to overflow when added, lets store
- the extra bit in mover. -- A special case exists when
- mantissa_bits is 31 (E_PRECISION). Then the first test cannot
- be trusted, as result is host-dependent, thus the second
- test. */
- if( smant == ((unsigned)(1<<(mantissa_bits+1))-1)
- || smant == (unsigned)-1 ) /* This is to catch E_PRECISION cases */
+ the extra bit in mover. */
+ if (smant == (1u << mantissa_bits << 1) - 1)
mover=1;
smant++;
}
/* Get the scaled one value */
- sone = (1 << (mantissa_bits));
+ sone = 1u << mantissa_bits;
/* The number may be unnormalised so renormalise it... */
if(mover)
@@ -527,7 +523,7 @@ tic4x_gen_to_words (FLONUM_TYPE flonum, LITTLENUM_TYPE *words, int precision)
else
{
/* Insert the exponent data into the word */
- sfract |= exponent << (mantissa_bits+1);
+ sfract |= (unsigned) exponent << (mantissa_bits + 1);
if (precision == S_PRECISION)
words[0] = sfract;
--
Alan Modra
Australia Development Lab, IBM
^ permalink raw reply [flat|nested] 2+ messages in thread
* tic4x-coff: ubsan: various shift UB
@ 2020-08-31 10:50 Alan Modra
0 siblings, 0 replies; 2+ messages in thread
From: Alan Modra @ 2020-08-31 10:50 UTC (permalink / raw)
To: binutils
* config/tc-tic4x.c (tic4x_gen_to_words): Rewrite mantissa
overflow test without UB. Avoid other UB shifts by making them
unsigned.
diff --git a/gas/config/tc-tic4x.c b/gas/config/tc-tic4x.c
index 1889b47128..4efb9d4ae5 100644
--- a/gas/config/tc-tic4x.c
+++ b/gas/config/tc-tic4x.c
@@ -453,8 +453,8 @@ tic4x_gen_to_words (FLONUM_TYPE flonum, LITTLENUM_TYPE *words, int precision)
/* Store the mantissa data into smant and the roundbit into rbit */
for (p = flonum.leader; p >= flonum.low && shift > -16; p--)
{
- tmp = shift >= 0 ? *p << shift : *p >> -shift;
- rbit = shift < 0 ? ((*p >> (-shift-1)) & 0x1) : 0;
+ tmp = shift >= 0 ? (unsigned) *p << shift : (unsigned) *p >> -shift;
+ rbit = shift < 0 ? (((unsigned) *p >> (-shift-1)) & 0x1) : 0;
smant |= tmp;
shift -= 16;
}
@@ -463,18 +463,14 @@ tic4x_gen_to_words (FLONUM_TYPE flonum, LITTLENUM_TYPE *words, int precision)
if(rbit)
{
/* If the mantissa is going to overflow when added, lets store
- the extra bit in mover. -- A special case exists when
- mantissa_bits is 31 (E_PRECISION). Then the first test cannot
- be trusted, as result is host-dependent, thus the second
- test. */
- if( smant == ((unsigned)(1<<(mantissa_bits+1))-1)
- || smant == (unsigned)-1 ) /* This is to catch E_PRECISION cases */
+ the extra bit in mover. */
+ if (smant == (1u << mantissa_bits << 1) - 1)
mover=1;
smant++;
}
/* Get the scaled one value */
- sone = (1 << (mantissa_bits));
+ sone = 1u << mantissa_bits;
/* The number may be unnormalised so renormalise it... */
if(mover)
@@ -527,7 +523,7 @@ tic4x_gen_to_words (FLONUM_TYPE flonum, LITTLENUM_TYPE *words, int precision)
else
{
/* Insert the exponent data into the word */
- sfract |= exponent << (mantissa_bits+1);
+ sfract |= (unsigned) exponent << (mantissa_bits + 1);
if (precision == S_PRECISION)
words[0] = sfract;
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2020-09-01 6:47 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-09-01 6:47 tic4x-coff: ubsan: various shift UB Alan Modra
-- strict thread matches above, loose matches on Subject: below --
2020-08-31 10:50 Alan Modra
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).