TAKE/part of offset gob! returns contents from head of gob!

[Siskin-Bot]

Submitted by: BrianH

When you TAKE/part from a gob!, it returns a block of sub-gobs of the length that you specified in the length argument. That number of gobs is removed from the parent gob at the position passed to the function.

However, the gobs returned from the block are the ones at the beginning of the sub-gob list, not the ones at the offset position passed to TAKE. That means that the gobs returned are not necessarily the ones removed, and some of the ones removed can be lost.

Note that TAKE without /part works correctly.

>> a: make gob! []
== make gob! [offset: 0x0 size: 100x100 alpha: 0]
>> repend a [make gob! [text: "1"] make gob! [text: "2"] make gob! [text: "3"]]
== make gob! [offset: 0x0 size: 100x100 alpha: 0]
>> b: take/part next a 1
== [make gob! [offset: 0x0 size: 100x100 alpha: 0 text: "1"]]
; should be [make gob! [offset: 0x0 size: 100x100 alpha: 0 text: "2"]]
>> a/1
== make gob! [offset: 0x0 size: 100x100 alpha: 0 text: "1"]
>> a/2
== make gob! [offset: 0x0 size: 100x100 alpha: 0 text: "3"]
>> same? first a first b
== true  ; should be false

---

^{Imported from: CureCode [ Version: alpha 110 Type: Bug Platform: All Category: Datatype Reproduce: Always Fixed-in:none ]
Imported from: metaeducation#1797}

Comments:

Rebolbot added Type.bug and Status.important on Jan 12, 2016

---

Hostilefork added on Jan 29, 2019

---

Hostilefork commented on Jan 29, 2019:

This is still broken in R3-Alpha, but now it's fixed in Ren-C. The method by which it is fixed is that while a bit-tweaked structure is used to store the GOB! itself (parent, owner, flags, size, offset, old size, old offset, alpha, type...), an ordinary BLOCK! is used to store the child panes. Hence GOB's TAKE is implemented by delegating to the ordinary implementation for TAKE. As tradeoffs go, this seems a fair one.

metaeducation/ren-c@7e4a507#diff-28cbe579c0d704de5a2764ddf110f75eR1136

---