GedMendby CaliberUtilities
← Back to GedMend

What GedMend will and won't do — and how to check

Software that touches your irreplaceable files should not ask to be believed. Here is what GedMend does not do, and how to check the things it says it does — on your own files, before paying anything.

Last updated 1 September 2026

Every developer says they test their software. That is worth nothing to you, because you cannot see the tests, and it is being said by the person selling you the thing.

So this page is built the other way round. It states plainly what GedMend does not do, and then hands you the means to check the rest yourself: everything except the final export is free and never expires, and every export prints arithmetic you can verify against files whose contents you already know. Neither requires taking our word for anything.

What GedMend will not do

This list is longer than most products publish. That is deliberate: the failure cases are what you actually need to know before trusting software with a file you cannot replace.

Deliberate limits

Things it is not

Formats and platform

A known defect, published rather than waited for

A GEDCOM 5.5 file written back as 5.5.1 can carry constructs that 5.5.1 removed. GedMend reads 5.5 but does not write it, so a 5.5 file comes back labelled 5.5.1. The trouble is that 5.5.1 withdrew several things 5.5 allowed — a photograph embedded in the file itself, a media format recorded on the object rather than on the file, an age on an engagement, and a few date forms. Those lines are carried through faithfully, so nothing of yours is lost; what is wrong is the label. A strict validator will report them, and it will be reporting lines GedMend did not write.

Measured, on public test files, with the free validator you can run yourself at ged-inline.org. Our 5.5.1 output of Heiner Eichmann’s allged.ged carries nineteen such lines that the input is not reported for: six media formats and six notes recorded on a multimedia object, and seven file references with no format — which is every file reference in the file. If you run that validator yourself you will see five of each and no more: it suppresses long runs of identical messages, so its report shows fifteen. The nineteen comes from counting the output directly. Every one of those lines came from the input unchanged; what changed is the version they are being judged against. The TGC551LF torture file adds a status without a date, dates written B.C., and LDS status values, in the same shape.

GedMend now tells you at export when your file contains such constructs, and names them — for allged.ged it reports “this file uses a media format on the record (OBJE.FORM) and a note on the multimedia record (OBJE.NOTE)”. The proper fix is a converter, and it is not written.

Check it yourself, on your own files

Everything except the final export is free, with no time limit and nothing to expire. That is not a trial gimmick — it is what makes the paragraphs above checkable. Load your two real files, work through the matches, resolve the conflicts, read the change report, and decide whether we deserve the export fee after you have seen what we would write. Your originals are read-only throughout, so there is nothing to undo.

Two things are worth checking while you are in there, both against files whose contents you already know.

The count equation. Before anything is written, the preview shows: people in File A, plus people in File B, minus the pairs you confirmed, equals people out. Every confirmed pair removes exactly one duplicate, so the arithmetic comes out exactly or not at all. The same check runs for families. If it does not balance, GedMend says so rather than writing the file quietly.

The citation arithmetic. Every export writes a change report — in both HTML and plain text — giving per-person figures like Citations: A=2 B=3 merged=4. Add the inputs, subtract the ones that folded because they were identical, and compare. A merged count lower than the larger of the two inputs would mean evidence was discarded; that has a name, and it is a bug.

That headline figure counts every citation on the person, including the ones attached to a birth, marriage or death rather than to the person directly — which is how most programs write them. A second figure in brackets gives the count attached to the person alone. When the two differ, it is because your evidence hangs off the events, and nothing is wrong.

This is a better test than any prepared example could be, because it is your tree: you already know which two people are the same person and which source you cited twice. That is a stronger form of assurance than any statement on this page.

What GedMend will do

Stated compactly, since the rest of the site covers it in detail: