Re: [Nasm-devel] issues from the forum...

[Cyrill G. <gor...@gm...>]

On 05/22/2011 09:07 PM, H. Peter Anvin wrote:
> On 05/22/2011 07:10 AM, Cyrill Gorcunov wrote:
>>
>> Well, indeed we better should have there expression evaluator called, this would allow us
>> to write say
>>
>> 	align 1+7
>>
>> or something like this. I'll take a look, but can;t guarantee i do this that soon :(
>> Probably on next week.
>>
> 
> Yes, that would definitely be the better option.
> 
> 	-hpa
> 

OK, this is the patch I have in my queue atm. Please review. I also have some cleanup
series for 2.09.xx, hope I manage to merge all this during the week.

    Cyrill
---
From: Cyrill Gorcunov <gor...@gm...>
Date: Mon, 23 May 2011 23:50:03 +0400
Subject: [PATCH] nassm.c: Use evaluate for section alignment

This allow us to write the whole expressions
on section alignments, such as

	align 0xa+6

or whatever math. Should be a way more convenient
than hardnumbers scheme we had.

Reported-by: Frank Kotler <fbk...@zy...>
Signed-off-by: Cyrill Gorcunov <gor...@gm...>
---
 nasm.c |   20 ++++++++++++++++----
 1 files changed, 16 insertions(+), 4 deletions(-)

diff --git a/nasm.c b/nasm.c
index 36be46e..9baec18 100644
--- a/nasm.c
+++ b/nasm.c
@@ -1248,10 +1248,22 @@ static void assemble_file(char *fname, StrList **depend_ptr)
                     }
                     break;
                 case D_SECTALIGN:        /* [SECTALIGN n] */
-                    {
-                        if (*value) {
-                            unsigned int align = atoi(value);
-                            if (!is_power2(align)) {
+                    if (*value) {
+                        stdscan_reset();
+                        stdscan_set(value);
+                        tokval.t_type = TOKEN_INVALID;
+                        e = evaluate(stdscan, NULL, &tokval, NULL, pass2, nasm_error, NULL);
+                        if (e) {
+                            unsigned int align = (unsigned int)e->value;
+                            if ((uint64_t)e->value > 0x7fffffff) {
+                                /*
+                                 * FIXME: Please make some sane message here
+                                 * ofmt should have some 'check' method which
+                                 * would report segment alignment bounds.
+                                 */
+                                nasm_error(ERR_FATAL,
+                                           "incorrect segment alignment `%s'", value);
+                            } else if (!is_power2(align)) {
                                 nasm_error(ERR_NONFATAL,
                                            "segment alignment `%s' is not power of two",
                                             value);
-- 
1.7.5.1