- s-nail-14_5_2-sort.patch
<http://downloads.sourceforge.net/project/s-nail/s-nail-14_5_2-sort.patch>
Reverses (sort(),thread(): use srelax()!, 2014-01-18, [a9b67e9]),
which was a hasty commit of an untested diff that i've added few
minutes beforehand:
As a rather careless last-minute change i've added string relaxation
to threaded and sorted display, but it's really one more step towards
lowering memory pressure -- i couldn't resist [a9b67e9] after seeing
An alternative, forward-heading patch that keeps string relaxation has
been pushed to [master] as [5e75529] and is also available as
s-nail-14_5_2-sort-alt.patch.
I'll attach both patches; git(1) users should update their [master].
Sorry, but hey, memory relaxation is pretty damn cool. Especially
with large boxes. (I must have been blinded. Or whatever.)
--steffen
s-nail-14_5_2-sort-alt.patch, 2014-01-30:
Apply:
$ cd s-nail-14.5.2
$ patch -bu < s-nail-14_5_2-sort-alt.patch
Description:
An alternative working thesis for the bugfix s-nail-14_5_2-sort.patch
(2014-01-30).
It instead uses heap allocations, thus continues to relax the dope storage, and
has been pushed to [master].
diff --git a/thread.c b/thread.c
index 3ac5966..12af1fa 100644
--- a/thread.c
+++ b/thread.c
@@ -144,7 +144,7 @@ mlook(char *id, struct mitem *mt, struct message *mdata, int mprime)
mp = &mt[c];
}
if (mdata != NULL && mp->mi_id == NULL) {
- mp->mi_id = id;
+ mp->mi_id = sstrdup(id);
mp->mi_data = mdata;
mdata->m_idhash = ~h;
}
@@ -395,6 +395,10 @@ makethreads(struct message *m, long cnt, int nmail)
threadroot = interlink(m, cnt, nmail);
finalize(threadroot);
+
+ for (i = 0; i < mprime; ++i)
+ if (mt[i].mi_id != NULL)
+ free(mt[i].mi_id);
free(mt);
mb.mb_threaded = 1;
}
@@ -612,11 +616,11 @@ sort(void *vp)
case SORT_TO:
if ((cp = hfield1(method == SORT_FROM ?
"from" : "to", mp)) != NULL) {
- ms[n].ms_u.ms_char = showname ?
- realname(cp) : skin(cp);
+ ms[n].ms_u.ms_char = sstrdup(showname ?
+ realname(cp) : skin(cp));
makelow(ms[n].ms_u.ms_char);
} else
- ms[n].ms_u.ms_char = UNCONST("");
+ ms[n].ms_u.ms_char = sstrdup("");
break;
default:
case SORT_SUBJECT:
@@ -624,12 +628,12 @@ sort(void *vp)
in.s = cp;
in.l = strlen(in.s);
mime_fromhdr(&in, &out, TD_ICONV);
- ms[n].ms_u.ms_char =
- savestr(skipre(out.s));
+ ms[n].ms_u.ms_char = sstrdup(
+ skipre(out.s));
free(out.s);
makelow(ms[n].ms_u.ms_char);
} else
- ms[n].ms_u.ms_char = UNCONST("");
+ ms[n].ms_u.ms_char = sstrdup("");
break;
}
ms[n++].ms_n = i;
@@ -652,6 +656,17 @@ sort(void *vp)
threadroot = &message[0];
finalize(threadroot);
mb.mb_threaded = 2;
+
+ switch (method) {
+ case SORT_FROM:
+ case SORT_TO:
+ case SORT_SUBJECT:
+ for (i = 0; i < n; ++i)
+ free(ms[i].ms_u.ms_char);
+ /* FALLTHRU */
+ default:
+ break;
+ }
ac_free(ms);
return ((vp && vp != (void *)-1 && !inhook && ok_blook(header))
? headers(msgvec) : 0);
s-nail-14_5_2-sort.patch, 2014-01-30:
Apply:
$ cd s-nail-14.5.2
$ patch -bu < s-nail-14_5_2-sort.patch
Description:
Reverses (sort(),thread(): use srelax()!, 2014-01-18, [a9b67e9]), which
was a hasty commit of an untested diff that i've added few minutes
beforehand.
It was of course wrong.
The problem (no memory relaxation during entire sort operation) requires
a different approach (either only relax when we don't need to hold
dope storage (as for SORT_TO/SORT_FROM/SORT_SUBJECT/+) or use heap
memory for those allocations, then).
s-nail-14_5_2-sort-alt.patch is an alternative working thesis that uses
heap allocations, thus continues to relax the dope storage.
This version has been pushed to [master].
diff --git a/thread.c b/thread.c
index 3ac5966..9ec1ac9 100644
--- a/thread.c
+++ b/thread.c
@@ -361,8 +361,6 @@ makethreads(struct message *m, long cnt, int nmail)
return;
mprime = nextprime(cnt);
mt = scalloc(mprime, sizeof *mt);
-
- srelax_hold();
for (i = 0; i < cnt; i++) {
if ((m[i].m_flag&MHIDDEN) == 0) {
mlook(NULL, mt, &m[i], mprime);
@@ -376,7 +374,6 @@ makethreads(struct message *m, long cnt, int nmail)
m[i].m_level = 0;
if (!nmail && !(inhook&2))
m[i].m_collapsed = 0;
- srelax();
}
/*
* Most folders contain the eldest messages first. Traversing
@@ -387,12 +384,8 @@ makethreads(struct message *m, long cnt, int nmail)
* are replies to the one message, and are sorted such that
* youngest messages occur first.
*/
- for (i = cnt-1; i >= 0; i--) {
+ for (i = cnt-1; i >= 0; i--)
lookup(&m[i], mt, mprime);
- srelax();
- }
- srelax_rele();
-
threadroot = interlink(m, cnt, nmail);
finalize(threadroot);
free(mt);
@@ -572,8 +565,6 @@ sort(void *vp)
default:
break;
}
-
- srelax_hold();
for (n = 0, i = 0; i < msgCount; i++) {
mp = &message[i];
if ((mp->m_flag&MHIDDEN) == 0) {
@@ -637,10 +628,7 @@ sort(void *vp)
mp->m_child = mp->m_younger = mp->m_elder = mp->m_parent = NULL;
mp->m_level = 0;
mp->m_collapsed = 0;
- srelax();
}
- srelax_rele();
-
if (n > 0) {
qsort(ms, n, sizeof *ms, func);
threadroot = &message[ms[0].ms_n];
------------------------------------------------------------------------------
WatchGuard Dimension instantly turns raw network data into actionable
security intelligence. It gives you real-time visual feedback on key
security issues and trends. Skip the complicated setup - simply import
a virtual appliance and go from zero to informed in seconds.
http://pubads.g.doubleclick.net/gampad/clk?id=123612991&iu=/4140/ostg.clktrk
_______________________________________________
S-nail-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/s-nail-users
