A selection can remain flush with the canvas edge after Shrink because GIMP treats an edge-touching selection differently depending on the dialog option Selected areas continue outside the image. When the option is enabled, an edge that touches the image boundary is allowed to shrink inward. When it is disabled, that edge can remain at the image border. This concerns the selection’s geometry; it does not resize a layer or change image pixels. See the official Shrink command documentation.
Judgment and steps
The relevant boundary is the image canvas boundary, not necessarily the active layer boundary. A selection is the editable region, while a layer is the content container that may occupy only part of the canvas. For this question, first establish which sides of the selection actually touch the image edge.
- Create or preserve a selection that touches one or more canvas edges. Use a duplicate document or a reversible working state if the selection is important.
- Choose
Select→Shrink…, enter a modest amount such as10 px, and note whether the selection touches the top, right, bottom, or left image edge before confirming. - Run the comparison with Selected areas continue outside the image enabled. Record each touching side that visibly moves inward after Shrink.
- Recreate the same selection, use the same shrink amount, clear the option, and compare the same four sides. Do not change the layer, canvas size, selection shape, or amount between trials.
- Record the result by side. An edge that remains flush in the unchecked comparison is evidence of the option’s boundary behavior, not proof that Shrink failed globally.
The documentation describes Shrink as moving selection-edge points away from the nearest image edge toward the selection center. It also states that the option matters specifically when the selection runs along the image edge. Feathering is preserved, although its shape may change near corners or sharp curves. That is a property of the selection, not a pixel-editing operation.
Acceptance worksheet with new observations
The table below is a proposed worksheet using hypothetical observations, not a guarantee of exact dimensions or a substitute for the documented rule. The values illustrate how to record decisions rather than predict GIMP’s pixel-by-pixel result.
| Hypothetical setup | Option | Observed sides after 10 px Shrink | Decision |
|---|---|---|---|
| Selection touches left and top; right and bottom are inset | Enabled | Left and top move inward; inset sides also reduce | Treat the edge retreat as intentional |
| Same selection and amount | Disabled | Left and top remain flush; inset sides reduce | The edge behavior matches the unchecked setting |
| Selection is inset from all four sides | Enabled or disabled | All four sides move inward | The edge-extension option is not the deciding variable here |
| Selection touches a layer boundary but not the canvas | Either | Result follows the canvas-edge relationship, not the layer rectangle | Do not diagnose this as a layer-size change |
Acceptance should be stated per side: “left and top retreated” is more useful than “the selection still appears.” The marching outline may be difficult to judge at corners, so compare straight edge segments and keep the two trials otherwise identical.
Limits and the next step after failure
This option does not expand the canvas, enlarge the active layer, or alter image content. It only changes how Shrink handles selection edges coincident with the image boundary. A layer edge elsewhere in the document is therefore a separate observation and should not be used as evidence for this setting.
If switching the option produces no visible difference, verify that the selection truly touches the image canvas and that the same selection was recreated for both trials. Inspect straight sections rather than relying only on a corner. If every side is already inset, the documented edge condition is absent. If the selection is feathered, judge the selection’s boundary while allowing for the documented change in feather shape near curves. If the result remains unclear, reduce the comparison to one straight edge and one modest shrink amount before investigating unrelated layer or tool settings.
Three short questions
Does this change the image?
No. Shrink changes the selection region. Pixels remain unchanged until another operation uses that selection.
Does a touching layer edge prove the same problem?
No. The documented option refers to the image edge. A layer boundary is a different rectangle and needs separate diagnosis.
Why test all four sides?
Because a selection may touch only one edge, two adjacent edges, or several sides. Recording each side prevents a visible unchanged edge from being mistaken for a total Shrink failure.
For the exact control wording and its edge-specific condition, consult the official Shrink reference.
Sources and next step
This guide interprets public documentation and uses explicitly hypothetical acceptance examples, not application tests. Controls and availability can change; verify the cited manual against your installed version.
For work created in Imagild, carry the input requirements and acceptance checklist into the photo editor. The external software controls discussed above belong to their respective products.