(A) Fixes for the assertion failures described in issue #2 (see John Mainzer's last checkin message):

--src/H5PB.c: checks for size >= page size
--src/H5MF.c: disable/enable page buffering in H5MF_tidy_self_referential_fsm_hack()
--src/H5MFsection.c: call H5PB_remove_entry() for both raw/metadata pages in H5MF__sect_small_merge()
(B) Port and modify existing concurrent swmr tests to VFD SWMR. Also modify the following:
--remove flushes from VFD SWMR writer tests
--set Nreaders to 0 in test/testvfdswmr.sh.in to test for writers only
Please enter the commit message for your changes. Lines starting

1 files changed, 5 insertions, 5 deletions

diff --git a/src/H5MFsection.c b/src/H5MFsection.c
index 1d217a1..3b0a1c2 100644
--- a/src/H5MFsection.c
+++ b/src/H5MFsection.c
@@ -771,11 +771,11 @@ H5MF__sect_small_merge(H5FS_section_info_t **_sect1, H5FS_section_info_t *_sect2
         if(H5MF_xfree(udata->f, udata->alloc_type, (*sect1)->sect_info.addr, (*sect1)->sect_info.size) < 0)
             HGOTO_ERROR(H5E_RESOURCE, H5E_CANTFREE, FAIL, "can't free merged section")
 
-        /* Need to free possible metadata page in the PB cache */
-        /* This is in response to the data corruption bug from fheap.c with page buffering + page strategy */
-        /* Note: Large metadata page bypasses the PB cache */
-        /* Note: Update of raw data page (large or small sized) is handled by the PB cache */
-        if(udata->f->shared->pb_ptr != NULL && udata->alloc_type != H5FD_MEM_DRAW)
+        /* Need to free possible raw/metadata page in the page buffer.
+         * This is in response to the data corruption bug from test/fheap.c
+         * when page buffering + page aggregation strategy are used.
+         * Note: Large raw/metadata page bypasses the page buffer */
+        if(udata->f->shared->pb_ptr != NULL)
             if(H5PB_remove_entry(udata->f, (*sect1)->sect_info.addr) < 0)
                 HGOTO_ERROR(H5E_RESOURCE, H5E_CANTFREE, FAIL, "can't free merged section")