From: jose.kahan@w3.org
Date: Tue Nov 23 1999 - 07:35:55 CST
I just found this strange behavior on the latest HM out from CVS.
I don't know if it was there before.
When running hypermail on a file with a PDF and a .zip MIME attachments,
the attachments are decoded and stored correctly in the attach dir. However,
the links to these attachments were generated as if they were in-lined images,
<IMG ..> rather than external ones <A HREF= ..>.
I traced the problem so far to the the prefered_types option in the .hmrc
file.
In my .hmrc file I have this option:
prefered_types = text/plain text/html *
And no inline_types option (so I should use its default value).
When I have the ending "*" char, the set_inline_types variable takes
the value of the set_prefered_types one. That's why I see the links to the
attachments as <IMG>.
If I remove the ending "*" char, it works as expected. The set_inline_types
takes its default value.
On the contrary, if I let my options with the ending "*" char and add a
new one inline_types with the same value as the default one for this option,
it doesn't work.
I think the problem may come from the initialization of the default values
in setup.c:
=============================
case CFG_LIST:
if (defval) {
char *tpstr;
tpstr = strsav(defval);
*(struct hmlist **)cfg[i].value = NULL;
*(struct hmlist **)cfg[i].value =
(void *)add_list((struct hmlist *)cfg[i].value, tpstr);
free(tpstr);
}
else
*(struct hmlist **)cfg[i].value = NULL;
break;
================================
>From my debugging session, set_inline_types mysteriously has the same values
as set_prefered_types.
Both pointers are different:
(gdb) p set_prefered_types
$114 = (struct hmlist *) 0x4a638
(gdb) p set_inline_types
$115 = (struct hmlist *) 0x484b4
(gdb)
But both structures are linked (set_inline_types.next == set_prefered_types).
(gdb) p *set_inline_types
$117 = {val = 0x484b4 "", next = 0x4a638}
(gdb) p set_prefered_types
$113 = {val = 0x4b670 "image/gif", next = 0x4a648}
As both set_inline_types and set_prefered types are intialized to NULL,
this lines looks suspicious:
*(struct hmlist **)cfg[i].value = NULL;
As it looks like if we're trying to assing something to an unallocated
memory zone.
I'll try changing it to:
(struct hmlist **)cfg[i].value = NULL;
Is anyone more familiar with this code? Your expertise may save me debugging
time :)
In all cases, it looks like a real bug.
Thanks,
-Jose