[hypermail] Serious message skipping problem with hypermail

From: Jose Kahan <jose.kahan_at_w3.org_at_hypermail-project.org>
Date: Thu, 5 Jun 2003 16:56:05 +0200
Message-ID: <20030605145605.GA32236_at_inrialpes.fr>

Problem description

---

I noticed that in the latest hypermail, the messages in an archive don't start with the same message number, depending on which option you use. Starting with a brand new archive:

  hypermail -d dirname -u -m /path/to/mailbox

starts messages at number one (0001.html).

However if I invoke it as:

  hypermail -d dirname -m /path/to/mailbox

starts messages with number zero (0000.html).

This is a big problem, because it breaks down the archives. For example, I'm using the -u option to to add new messages to the archive as they arrive to the mailing list, on the fly.

However, if I want to rebuild the archives and use the second option, all my messages will skip by one.

Source of the problem

---

It happens in hypermail.c:593:

---

   /*

• Let's do it. */

    if (set_uselock)

        lock_archive(set_dir);

    if (set_increment) {

	int num_displayable;
	if (set_linkquotes)
	    replylist = NULL;
	num_displayable = loadoldheaders(set_dir);
	amount_old = max_msgnum + 1; /* counts gaps as messages */
	/* start numbering at this number */
		amount_new = num_displayable + parsemail(set_mbox,

use_stdin, set_readone, set_increment, set_dir, set_inlinehtml, amount_old);

---

max_msgnum gets incremented when calling loadoldheaders. However, it's initial value is 0. So, when doing the:

    mount_old = max_msgnum + 1; /* counts gaps as messages */

we automatically get one message skip if the archive was initially empty.. If we initialize max_msgnum to -1 before invoking loadoldheaders, then we don't have the skip anymore. And if there are any previous messages, max_msgnum gets updated correctly.

This problem doesn't happen when not using the -u option.

Proposed patches

---

1. Fix the problem.

  if (set_increment) {

	int num_displayable;
	if (set_linkquotes)
	    replylist = NULL;
        max_msgnum = -1;
  
        ^^^ just initialize it before calling loadoldheaders.

2. Add a backwards compatibility option, so that people who need to rebuild their archives can precise if they want them to start with 1 or 0.

Something like:

   start_num_is_0

And initialize max_msgnum accordingly (another patch is required when not using -u, to change max_msgnum accordingly. I tested it and it works ok too.)

---

Any comments?

In all cases, this is an awkward bug.

-jose Received on Thu 05 Jun 2003 03:59:26 PM GMT