When this is enabled for a mailbox rule, the specified Sieve script is executed for the message in the source mailbox during a "COPY" event.
This happens only after the Sieve script that is executed for the corresponding message in the destination mailbox finishes running successfully.

This already has the imapsieve transaction context associated with it before the sieve scripts are executed.
This currently cannot occur, but it will become relevant in subsequent commits.

imapsieve plugin: Don't open the user's personal script storage if no personal script is configured.

lib-sieve: multiscript: The execution of the discard script had an implicit "keep", rather than an implicit "discard".

lib-sieve: store action: Made static analyzer happier about tr_context argument for sieve_act_store_add_flags().

Added an assert ensuring it is not NULL.

This used strftime() directly and allocated memory from the heap before i_fatal().
It also used printf() to send the BYE message, which will only work right when the service is running stand-alone.

This adds a new callback to the Sieve script execution environment.

It can never be NULL after the "lib-storage: Always create
mail_save_context.dest_mail" core change.

lib-sieve: file storage: Amended the up-to-date timestamp comparison for on-disk binaries to include nanoseconds.

This will fix problems occurring when both binary and script are saved within the same second.
On older systems that have no support for nanoseconds in the stat() time stamps, this has no effect.

Otherwise, the mailbox name would never match, since matching is performed case-sensitively and Dovecot only returns the upper-cased "INBOX".
This change prevents confusing administrators that for some reason choose to use a different capitalization of that mailbox name.

Use eacces_error_get(), rather than just printing "Permission denied".

Use eacces_error_get(), rather than just printing "Permission denied".

path-util is a merge of realpath from Pigeonhole and abspath from core.

This is not currently used anywhere, so nothing actually broke.
This bug was exposed by the previous commit, causing a compiler error/warning on some compilers.

Applies to LDA/LMTP context in general, not only when sieve_before/sieve_after are used.
Action side-effects and the message snapshot would be lost at the final stage where the implicit keep is evaluated.
This happened because the keep action itself is not actually executed, but rather its presence is noted to determine whether more scripts need to be executed.
So, when finally execution of the actual keep action is due, i.e. when there are no more scripts in the sequence, it overrides the explicit keep from the last script.
This didn't take the side-effects and message snapshot into account.

The situation where no headers are found was not always handled correctly.

Used the following script:

C_FILES=`git ls-files *.c`
H_FILES=`git ls-files *.h`
for F in "$C_FILES $H_FILES"; do
        echo "$F"
  perl -p -i -e 's/safe_memset\(&\(?([^,]*)\)?,\s*0,\s*sizeof\(\g1\)\)/i_zero_safe(&$1)/g' $F
  perl -p -i -e 's/safe_memset\(([^,]*),\s*0,\s*sizeof\(\*\g1\)\)/i_zero_safe($1)/g' $F
  perl -p -i -e 's/memset\(&\(?([^,]*)\)?,\s*0,\s*sizeof\(\g1\)\)/i_zero(&$1)/g' $F
  perl -p -i -e 's/memset\(([^,]*),\s*0,\s*sizeof\(\*\g1\)\)/i_zero($1)/g' $F
done

Apparently, this causes GCC warnings.

Apparently, this causes GCC warnings.

In that case a transaction (delayed sync) is perfromed outside of the context of an IMAP command.

This avoids accidents with the array numbering being wrong.