Hello Stephen.
Stephen Isard wrote in
<[email protected]>:
|Here is a complete session where I
|
|1. Began a message with the word "Before"
|2. Entered ed as an external editor and added the line "During"
|3. Waited a long time
|4. Did a date stamp from within the editor, then saved and exited
|5. Got an IMAP write error
|6. Did ~p to see that the message only contained the "Before" line
Cannot reproduce this.
|7. Still within the message, did an external shell command to see that
| there was a file /tmp/s-nailblahblah with the same date stamp as
| when I left the editor.
|8. Still within the message, did an external cat command to see that the
| file in /tmp contained the "Before" and "During" lines.
|9. Abandoned the message with `x
|10. Tried to repeat the cat command on the /tmp file and found that
| it had been removed.
|
|So I am not accusing you of leaving files lying around in /tmp. But it
|appears that they aren't immediately deleted on return to the s-nail
|command line after an IMAP write error, when the content of the /tmp
|file hasn't been read back into the message buffer. Finishing the
|message appears to get the file cleaned away.
The compose mode is still active.
...
|s-nail: IMAP write error: error:00000000:lib(0):func(0):reason(0)
May i mention that it likely should not look like this for the
[master] branch that includes the patch for the error you
reported? Or where those errors washed away by terminal redraws?
...
|
|Before
|~!ls -gG /tmp/s-nail*
|-rw------- 1 14 Aug 30 16:12 /tmp/s-nail-edbaseUjFRoCne
|!
|(continue)
|~!cat /tmp/s-nail-edbaseUjFRoCne
|Before
|During
|!
|(continue)
|~x
|s-nail: IMAP write error: Bad file descriptor
That all appears strange, Stephen. We may have two files active
once you edit, but if the editor returns with a successful exit
status we replace the old with the new one, throwing away the old
one; otherwise we throw away the new one. That is all the code
there is.
|ERROR# & !cat /tmp/s-nail-edbaseUjFRoCne
|cat: /tmp/s-nail-edbaseUjFRoCne: No such file or directory
So you could apply the patch below to the [master] branch.
My only idea would ... is that i do not have one. Even if the
file system does not update the modification time, the file size
surely failed.
diff --git a/src/mx/cmd-edit.c b/src/mx/cmd-edit.c
index acd9d4e3ed..f7d1ead002 100644
--- a/src/mx/cmd-edit.c
+++ b/src/mx/cmd-edit.c
@@ -242,20 +242,32 @@ jetempo:
cc.cc_args[0] = fstcp->fstc_filename;
}
+fprintf(stderr, "PRE CHILD RUN cmd <%s>\n", cc.cc_cmd);
if(!mx_child_run(&cc) || cc.cc_exit_status != 0)
+{
+fprintf(stderr, "BAD CHILD EXIT status %d err %d\n",
cc.cc_exit_status,cc.cc_error);
goto jleave;
+}
+fprintf(stderr, "GOOD CHILD EXIT\n");
/* If in read only mode or file unchanged, just remove the editor temporary
* and return. Otherwise switch to new file */
if(viored != '|'){
if(readonly)
+{
+fprintf(stderr, "DISCARD EDIT, readonly\n");
goto jleave;
+}
if(stat(fstcp->fstc_filename, &statb) == -1){
+fprintf(stderr, "DISCARD EDIT, stat error\n");
n_perr(fstcp->fstc_filename, 0);
goto jleave;
}
if(modtime == statb.st_mtime && modsize == statb.st_size)
+{
+fprintf(stderr, "DISCARD EDIT, time unchanged\n");
goto jleave;
+}
}
if((nf = mx_fs_open(fstcp->fstc_filename, "r+")) == NIL)
--steffen
|
|Der Kragenbaer, The moon bear,
|der holt sich munter he cheerfully and one by one
|einen nach dem anderen runter wa.ks himself off
|(By Robert Gernhardt)
