Hallo, attached find a very conservative patch against your patched cmd-load-buffer.c which brings the number of calls to realloc() from n=(final buffer size) to n<log_2(final buffer size). Fulvio
tmux.diff
Description: Binary data
At Thu, 18 Feb 2010 23:20:19 +0000,
Nicholas Marriott wrote:
>
> Hi
>
> Thanks for the diff.
>
> Tiago Cunha who originally wrote the load-buffer command sent me an update to
> do this as well which is quite similar to yours but a bit simpler, so we will
> probably use it. I've attached it below as well if you want to take a look.
>
> You can't just use stdin because the loadb command is running in the server so
> stdin doesn't exist.
>
> At the moment the client doesn't send the stdin fd up to the client unless it
> is a tty, we could change that by removing the isatty() check in client.c.
>
> There may be a couple of things that need checked after that, eg would
> probably
> need an isatty() check in tty_open() and maybe elsewhere.
>
> Once that is done, the load-buffer command could check that ctx->cmdclient is
> non-NULL. If it is then the command was run from the command-line and the tty
> fd will be in c->tty.fd.
>
> Then it should be able to call isatty() itself, then freopen() on that fd if
> it
> isn't a tty, and read it.
>
> It might be nice to have stdin available for other things as well so this
> could
> be worth looking into.
>
>
> --- cmd-load-buffer.c.orig 2010-02-18 21:44:50.539052646 +0000
> +++ cmd-load-buffer.c 2010-02-18 21:50:52.395061286 +0000
> @@ -16,13 +16,10 @@
> * OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
> */
>
> -#include <sys/types.h>
> -#include <sys/stat.h>
> -
> #include <errno.h>
> +#include <stdio.h>
> #include <stdlib.h>
> #include <string.h>
> -#include <unistd.h>
>
> #include "tmux.h"
>
> @@ -48,11 +45,11 @@
> {
> struct cmd_buffer_data *data = self->data;
> struct session *s;
> - struct stat sb;
> FILE *f;
> - char *pdata = NULL;
> + char *pdata, *new_pdata;
> size_t psize;
> u_int limit;
> + int ch;
>
> if ((s = cmd_find_session(ctx, data->target)) == NULL)
> return (-1);
> @@ -62,29 +59,23 @@
> return (-1);
> }
>
> - if (fstat(fileno(f), &sb) < 0) {
> - ctx->error(ctx, "%s: %s", data->arg, strerror(errno));
> - goto error;
> + pdata = NULL;
> + psize = 0;
> + while ((ch = getc(f)) != EOF) {
> + /* Do not let the server die due to memory exhaustion. */
> + if ((new_pdata = realloc(pdata, psize + 2)) == NULL) {
> + ctx->error(ctx, "realloc error: %s", strerror(errno));
> + goto error;
> + }
> + pdata = new_pdata;
> + pdata[psize++] = ch;
> }
> - if (sb.st_size <= 0 || (uintmax_t) sb.st_size > SIZE_MAX) {
> - ctx->error(ctx, "%s: file empty or too large", data->arg);
> - goto error;
> - }
> - psize = (size_t) sb.st_size;
> -
> - /*
> - * We don't want to die due to memory exhaustion, hence xmalloc can't
> - * be used here.
> - */
> - if ((pdata = malloc(psize)) == NULL) {
> - ctx->error(ctx, "malloc error: %s", strerror(errno));
> - goto error;
> - }
> -
> - if (fread(pdata, 1, psize, f) != psize) {
> - ctx->error(ctx, "%s: fread error", data->arg);
> + if (ferror(f)) {
> + ctx->error(ctx, "%s: %s", data->arg, strerror(errno));
> goto error;
> }
> + if (pdata != NULL)
> + pdata[psize] = '\0';
>
> fclose(f);
>
------------------------------------------------------------------------------ Download Intel® Parallel Studio Eval Try the new software tools for yourself. Speed compiling, find bugs proactively, and fine-tune applications for parallel performance. See why Intel Parallel Studio got high marks during beta. http://p.sf.net/sfu/intel-sw-dev
_______________________________________________ tmux-users mailing list tmux-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/tmux-users
