Turning off mergeinfo calculations against a pre-1.5 repos

[Daniel Rall <dl...@collab.net>]

Mike, was this along the lines of what you were thinking?  On IRC, you
mentioned adjusting the ignore_ancestry flag directly, which we could
do instead of adding another field (unsure whether doing so preserve
the semantics).

- Dan


On Mon, 26 Nov 2007, dlr@tigris.org wrote:
...
> * subversion/libsvn_client/merge.c
>   Include svn_ra.h.
>   (merge_cmd_baton_t): Add new mergeinfo_capable field, indicating
>    whether the merge source repository is capable of Merge Tracking.
>   (mergeinfo_behavior): Add APR_INLINE to declaration.  Test
>    *MERGE_B->mergeinfo_capable when setting both *HONOR_MERGEINFO and
>    *RECORD_MERGEINFO.
>   (do_merge): Set *merge_cmd_baton.mergeinfo_capable using the new
>    svn_ra_has_capability() API.
...
> --- trunk/subversion/libsvn_client/merge.c	(original)
> +++ trunk/subversion/libsvn_client/merge.c	Mon Nov 26 16:17:12 2007
> @@ -43,6 +43,7 @@
>  #include "svn_props.h"
>  #include "svn_time.h"
>  #include "svn_sorts.h"
> +#include "svn_ra.h"
>  #include "client.h"
>  #include "mergeinfo.h"
>  #include <assert.h>
> @@ -170,6 +171,8 @@
>                                           is the same repository as the
>                                           target.  Defaults to FALSE if DRY_RUN
>                                           is TRUE.*/
> +  svn_boolean_t *mergeinfo_capable;   /* Whether the merge source repository
> +                                         is capable of Merge Tracking. */
>    svn_boolean_t ignore_ancestry;      /* Are we ignoring ancestry (and by
>                                           extension, mergeinfo)?  FALSE if
>                                           SOURCES_RELATED is FALSE. */
> @@ -2149,17 +2152,20 @@
>  
>  /* Set *HONOR_MERGEINFO and *RECORD_MERGEINFO (if non-NULL)
>     appropriately for MERGE_B. */
> -static void
> +static APR_INLINE void
>  mergeinfo_behavior(svn_boolean_t *honor_mergeinfo,
>                     svn_boolean_t *record_mergeinfo,
>                     merge_cmd_baton_t *merge_b)
>  {
>    if (honor_mergeinfo)
> -    *honor_mergeinfo = (merge_b->sources_related
> +    *honor_mergeinfo = (*merge_b->mergeinfo_capable
> +                        && merge_b->sources_related
>                          && merge_b->same_repos
>                          && (! merge_b->ignore_ancestry));
> +
>    if (record_mergeinfo)
> -    *record_mergeinfo = (merge_b->sources_related
> +    *record_mergeinfo = (*merge_b->mergeinfo_capable
> +                         && merge_b->sources_related
>                           && merge_b->same_repos
>                           && (! merge_b->dry_run));
>  }
> @@ -4299,6 +4305,7 @@
>    svn_config_t *cfg;
>    const char *diff3_cmd;
>    int i;
> +  svn_boolean_t mergeinfo_capable = FALSE;
>  
>    /* If this is a dry-run record-only merge, there's nothing to do. */
>    if (record_only && dry_run)
> @@ -4330,6 +4337,7 @@
>    merge_cmd_baton.record_only = record_only;
>    merge_cmd_baton.ignore_ancestry = ignore_ancestry;
>    merge_cmd_baton.same_repos = same_repos;
> +  merge_cmd_baton.mergeinfo_capable = NULL;
>    merge_cmd_baton.sources_related = sources_related;
>    merge_cmd_baton.ctx = ctx;
>    merge_cmd_baton.target_missing_child = FALSE;
> @@ -4394,6 +4402,15 @@
>        merge_cmd_baton.ra_session1 = ra_session1;
>        merge_cmd_baton.ra_session2 = ra_session2;
>  
> +      /* Populate the portions of the merge context baton that require
> +         an RA session to set, but shouldn't be reset for each iteration. */
> +      if (merge_cmd_baton.mergeinfo_capable == NULL)
> +        {
> +          merge_cmd_baton.mergeinfo_capable = &mergeinfo_capable;
> +          SVN_ERR(svn_ra_has_capability(ra_session1, &mergeinfo_capable,
> +                                        SVN_RA_CAPABILITY_MERGEINFO, pool));
> +        }
> +
>        /* If this is a record-only merge and our sources are from the
>           same repository as our target, just do the record and move on. */
>        if (same_repos && record_only)

Re: Turning off mergeinfo calculations against a pre-1.5 repos

["C. Michael Pilato" <cm...@collab.net>]

Daniel Rall wrote:
> Mike, was this along the lines of what you were thinking?  On IRC, you
> mentioned adjusting the ignore_ancestry flag directly, which we could
> do instead of adding another field (unsure whether doing so preserve
> the semantics).

What you did was more-or-less what I had in mind.  After we spoke on IRC, I
later realized that we should *not* overload ignore_ancestry in this case.

See my commit review for comments about the details of your change, though.

-- 
C. Michael Pilato <cm...@collab.net>
CollabNet   <>   www.collab.net   <>   Distributed Development On Demand