[ADD] migration: look at a COW copy before agreeing to give it up
The prompt asked whether to neutralize copies without showing what they hold.
Answering meant giving up a customization sight unseen — often three lines, an
id and a container width, sometimes a whole page, and nothing told them apart.
« v » now shows both halves of the question. What the copy changed, diffed
against the module view it shadows: on the database at hand, 5 lines added and
3 removed, two CSS anchors and a container width. And why it breaks, by
showing the declaration in each version — portal.frontend_layout is a
standalone template in 12.0 and inheritance specs in 13.0, which is the whole
explanation. « w » is the same two, full screen, space to switch.
The current version comes from ir_module_module, not .odoo-version: the
checkout is switched to the target before this runs, so reading the file
compared the target with itself and printed the same declaration twice. That
is what it did until it was run against a real migration.
--list says which copies a past neutralization put aside. A renamed key is
invisible in the interface, so without it the only trace was remembering.
Checked on the VM mid-migration: both views on view 2670, the screen builds
headless and toggles, --list reports and reports nothing when there is
nothing. A missing database now exits 2 with a message instead of a traceback.
--- FR ---
L'invite demandait de neutraliser des copies sans montrer ce qu'elles
contiennent. Répondre revenait à renoncer à une personnalisation sans l'avoir
vue — souvent trois lignes, un id et une largeur de conteneur, parfois une
page entière, et rien ne les distinguait.
« v » montre désormais les deux moitiés de la question. Ce que la copie a
changé, comparé à la vue de module qu'elle masque : sur la base en cours, 5
lignes ajoutées et 3 retirées, deux ancres CSS et une largeur. Et pourquoi ça
casse, en affichant la déclaration dans chaque version —
portal.frontend_layout est un gabarit autonome en 12.0 et des consignes
d'héritage en 13.0, ce qui est toute l'explication. « w » donne les deux en
plein écran, espace pour basculer.
La version courante vient d'ir_module_module, pas de .odoo-version : le
checkout est basculé sur la cible avant cette étape, donc lire le fichier
comparait la cible avec elle-même et affichait deux fois la même déclaration.
C'est ce qu'il faisait jusqu'à l'essai sur une vraie migration.
--list dit quelles copies une neutralisation passée a mises de côté. Une clé
renommée est invisible dans l'interface ; sans cela, la seule trace était de
s'en souvenir.
Vérifié sur la VM en cours de migration : les deux vues sur la vue 2670,
l'écran se construit sans terminal et bascule, --list rapporte et ne rapporte
rien quand il n'y a rien. Une base absente sort en 2 avec un message plutôt
qu'une trace d'appel.
Assisted-by: Claude Opus 5
2026-08-11 23:49:05 -04:00
|
|
|
#!/usr/bin/env python3
|
|
|
|
|
# © 2021-2026 TechnoLibre (http://www.technolibre.ca)
|
|
|
|
|
# License AGPL-3.0 or later (http://www.gnu.org/licenses/agpl)
|
|
|
|
|
|
|
|
|
|
"""Show what a website COW copy holds, and why the next version breaks it.
|
|
|
|
|
|
|
|
|
|
Read-only. Answers the two questions someone asks before neutralizing a copy,
|
|
|
|
|
and neither is answerable from the warning alone:
|
|
|
|
|
|
|
|
|
|
**What do I lose?** The copy is compared with the module view it shadows. That
|
|
|
|
|
is the customization someone made, and often it is three lines — an id, a
|
|
|
|
|
container width — for which nobody would hold up a migration.
|
|
|
|
|
|
|
|
|
|
**Why does it break?** The module declaration is shown in the current version
|
|
|
|
|
and in the target. The pair is the whole explanation: a template declared
|
|
|
|
|
without ``inherit_id`` is a standalone document, one declared with it must
|
|
|
|
|
hold inheritance specs. A copy frozen in the first shape cannot be applied in
|
|
|
|
|
the second, and Odoo stops on « cannot be located in parent view ».
|
|
|
|
|
|
|
|
|
|
Nothing here writes. Neutralizing is ``neutralize_cow_views.py --apply``, and
|
|
|
|
|
undoing it is ``--restore``.
|
|
|
|
|
"""
|
|
|
|
|
|
|
|
|
|
import argparse
|
|
|
|
|
import difflib
|
|
|
|
|
import json
|
|
|
|
|
import os
|
|
|
|
|
import re
|
|
|
|
|
import subprocess
|
|
|
|
|
import sys
|
|
|
|
|
|
|
|
|
|
sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
|
|
|
|
|
|
|
|
|
|
from check_cow_views import analyse, find_module_dir # noqa: E402
|
|
|
|
|
|
|
|
|
|
# The declaration of a template spans a few lines; showing the opening tag and
|
|
|
|
|
# what follows is enough to see its shape, and short enough to compare two
|
|
|
|
|
# versions side by side without scrolling.
|
|
|
|
|
DECL_LINES = 4
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def run_psql(database, sql):
|
|
|
|
|
"""Run a statement and return stdout, raising on failure."""
|
|
|
|
|
result = subprocess.run(
|
|
|
|
|
["psql", "-X", "-w", "-d", database, "-tAc", sql],
|
|
|
|
|
capture_output=True,
|
|
|
|
|
text=True,
|
|
|
|
|
)
|
|
|
|
|
if result.returncode:
|
|
|
|
|
raise RuntimeError(
|
|
|
|
|
f"Query failed on '{database}': {result.stderr.strip()}"
|
|
|
|
|
)
|
|
|
|
|
return result.stdout
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def unwrap_arch(value):
|
|
|
|
|
"""The arch as text, whatever the column type.
|
|
|
|
|
|
|
|
|
|
``arch_db`` is text up to 15.0 and jsonb from 16.0. The same unwrapping as
|
|
|
|
|
the sibling scripts, kept identical on purpose: a second reading of the
|
|
|
|
|
same column that differs on an edge case would report a drift nobody made.
|
|
|
|
|
"""
|
|
|
|
|
text = (value or "").strip()
|
|
|
|
|
if text.startswith("{") and '"' in text:
|
|
|
|
|
try:
|
|
|
|
|
data = json.loads(text)
|
|
|
|
|
except ValueError:
|
|
|
|
|
return value or ""
|
|
|
|
|
if isinstance(data, dict) and data:
|
|
|
|
|
for lang in ("en_US", *sorted(data)):
|
|
|
|
|
if isinstance(data.get(lang), str):
|
|
|
|
|
return data[lang]
|
|
|
|
|
return value or ""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def fetch_arch(database, view_id):
|
|
|
|
|
"""The stored arch of one view."""
|
|
|
|
|
return unwrap_arch(
|
|
|
|
|
run_psql(
|
|
|
|
|
database,
|
|
|
|
|
f"SELECT arch_db::text FROM ir_ui_view WHERE id = {int(view_id)};",
|
|
|
|
|
)
|
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def fetch_module_view(database, key):
|
|
|
|
|
"""(id, arch) of the module view a copy shadows, or (None, '').
|
|
|
|
|
|
|
|
|
|
The module view is the one carrying that key WITHOUT a website: that is
|
|
|
|
|
exactly the pairing Odoo itself makes, and the reason renaming a key is
|
|
|
|
|
enough to unpair a copy.
|
|
|
|
|
"""
|
|
|
|
|
safe = key.replace("'", "''")
|
|
|
|
|
out = run_psql(
|
|
|
|
|
database,
|
|
|
|
|
"SELECT id, arch_db::text FROM ir_ui_view"
|
|
|
|
|
f" WHERE key = '{safe}' AND website_id IS NULL"
|
|
|
|
|
" ORDER BY id LIMIT 1;",
|
|
|
|
|
).strip()
|
|
|
|
|
if not out:
|
|
|
|
|
return None, ""
|
|
|
|
|
view_id, _, arch = out.partition("|")
|
|
|
|
|
return int(view_id), unwrap_arch(arch)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def declaration(version_dir, key):
|
|
|
|
|
"""(path, snippet) of how the sources of one version declare this key.
|
|
|
|
|
|
|
|
|
|
Returns None when the module or the node is not found — a fact worth
|
|
|
|
|
showing as such, since « the module no longer declares it » is itself a
|
|
|
|
|
reason a copy breaks.
|
|
|
|
|
"""
|
|
|
|
|
module_name, _, template_id = key.partition(".")
|
|
|
|
|
if not template_id:
|
|
|
|
|
return None
|
|
|
|
|
module_dir = find_module_dir(version_dir, module_name)
|
|
|
|
|
if module_dir is None:
|
|
|
|
|
return None
|
|
|
|
|
try:
|
|
|
|
|
from lxml import etree
|
|
|
|
|
except ImportError:
|
|
|
|
|
return None
|
|
|
|
|
|
|
|
|
|
pattern = os.path.join(module_dir, "**", "*.xml")
|
|
|
|
|
import glob
|
|
|
|
|
|
|
|
|
|
for path in sorted(glob.glob(pattern, recursive=True)):
|
|
|
|
|
try:
|
|
|
|
|
tree = etree.parse(path)
|
|
|
|
|
except etree.XMLSyntaxError:
|
|
|
|
|
continue
|
|
|
|
|
for element in tree.getroot().iter():
|
|
|
|
|
if element.get("id") != template_id:
|
|
|
|
|
continue
|
|
|
|
|
if element.tag not in ("template", "record"):
|
|
|
|
|
continue
|
|
|
|
|
# The source line, not a re-serialization: what the file actually
|
|
|
|
|
# says is what a reader will grep for.
|
|
|
|
|
with open(path, "r", encoding="utf-8", errors="replace") as handle:
|
|
|
|
|
lines = handle.read().splitlines()
|
|
|
|
|
start = max((element.sourceline or 1) - 1, 0)
|
|
|
|
|
return path, "\n".join(lines[start : start + DECL_LINES])
|
|
|
|
|
return None
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def database_version_dir(database):
|
|
|
|
|
"""The odoo<x>.0 directory matching what the DATABASE says it is.
|
|
|
|
|
|
|
|
|
|
Not `.odoo-version`: the checkout is switched to the TARGET before this
|
|
|
|
|
runs, so reading it would compare the target with itself and show the same
|
|
|
|
|
declaration twice — which is what it did until this was measured on a real
|
|
|
|
|
migration. The database, at that moment, is still on the previous version
|
|
|
|
|
and says so in ir_module_module.
|
|
|
|
|
"""
|
|
|
|
|
try:
|
|
|
|
|
out = run_psql(
|
|
|
|
|
database,
|
|
|
|
|
"SELECT latest_version FROM ir_module_module"
|
|
|
|
|
" WHERE name = 'base';",
|
|
|
|
|
).strip()
|
|
|
|
|
except RuntimeError:
|
|
|
|
|
return None
|
|
|
|
|
if not out:
|
|
|
|
|
return None
|
|
|
|
|
parts = out.split(".")
|
|
|
|
|
if len(parts) < 2:
|
|
|
|
|
return None
|
|
|
|
|
return f"odoo{parts[0]}.{parts[1]}"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def collect(database, target_version, current_version=None):
|
|
|
|
|
"""Everything needed to judge each at-risk copy. No writes."""
|
|
|
|
|
current_version = current_version or database_version_dir(database)
|
|
|
|
|
lst_at_risk, _, _ = analyse(database, target_version)
|
|
|
|
|
lst_finding = []
|
|
|
|
|
for view_id, key, mode, target_mode, website_id, reason in lst_at_risk:
|
|
|
|
|
module_id, module_arch = fetch_module_view(database, key)
|
|
|
|
|
lst_finding.append(
|
|
|
|
|
{
|
|
|
|
|
"id": view_id,
|
|
|
|
|
"key": key,
|
|
|
|
|
"mode": mode,
|
|
|
|
|
"target_mode": target_mode,
|
|
|
|
|
"website_id": website_id,
|
|
|
|
|
"reason": reason,
|
|
|
|
|
"copy_arch": fetch_arch(database, view_id),
|
|
|
|
|
"module_id": module_id,
|
|
|
|
|
"module_arch": module_arch,
|
|
|
|
|
"decl_current": (
|
|
|
|
|
declaration(current_version, key)
|
|
|
|
|
if current_version
|
|
|
|
|
else None
|
|
|
|
|
),
|
|
|
|
|
"decl_target": declaration(target_version, key),
|
|
|
|
|
"current_version": current_version,
|
|
|
|
|
"target_version": target_version,
|
|
|
|
|
}
|
|
|
|
|
)
|
|
|
|
|
return lst_finding
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def render_diff(finding):
|
|
|
|
|
"""What the copy changed, compared with the module view it shadows."""
|
|
|
|
|
lines = [
|
|
|
|
|
f"── id={finding['id']} {finding['key']}"
|
|
|
|
|
f" (website={finding['website_id']}) ──",
|
|
|
|
|
"",
|
|
|
|
|
]
|
|
|
|
|
if finding["module_id"] is None:
|
|
|
|
|
lines += [
|
|
|
|
|
" No module view carries this key, so there is nothing to compare",
|
|
|
|
|
" against: this copy is a page made in the website editor.",
|
|
|
|
|
]
|
|
|
|
|
return "\n".join(lines)
|
|
|
|
|
|
|
|
|
|
left = finding["module_arch"].splitlines()
|
|
|
|
|
right = finding["copy_arch"].splitlines()
|
|
|
|
|
diff = [
|
|
|
|
|
line
|
|
|
|
|
for line in difflib.unified_diff(
|
|
|
|
|
left,
|
|
|
|
|
right,
|
|
|
|
|
fromfile=f"module id={finding['module_id']}",
|
|
|
|
|
tofile=f"copy id={finding['id']}",
|
|
|
|
|
lineterm="",
|
|
|
|
|
n=1,
|
|
|
|
|
)
|
|
|
|
|
]
|
|
|
|
|
if len(diff) <= 2:
|
|
|
|
|
lines.append(" The copy is identical to the module view.")
|
|
|
|
|
return "\n".join(lines)
|
|
|
|
|
lines += [f" {line}" for line in diff]
|
|
|
|
|
n_plus = sum(
|
|
|
|
|
1 for x in diff if x.startswith("+") and not x.startswith("+++")
|
|
|
|
|
)
|
|
|
|
|
n_minus = sum(
|
|
|
|
|
1 for x in diff if x.startswith("-") and not x.startswith("---")
|
|
|
|
|
)
|
|
|
|
|
lines += [
|
|
|
|
|
"",
|
|
|
|
|
f" {n_plus} line(s) added, {n_minus} removed — this is what"
|
|
|
|
|
" neutralizing gives up.",
|
|
|
|
|
]
|
|
|
|
|
return "\n".join(lines)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def render_shape(finding):
|
|
|
|
|
"""Why it breaks: the declaration in each version, side by side."""
|
|
|
|
|
lines = [
|
|
|
|
|
f"── id={finding['id']} {finding['key']} ──",
|
|
|
|
|
"",
|
|
|
|
|
f" {finding['reason']}",
|
|
|
|
|
"",
|
|
|
|
|
]
|
|
|
|
|
for label, decl in (
|
|
|
|
|
(finding["current_version"], finding["decl_current"]),
|
|
|
|
|
(finding["target_version"], finding["decl_target"]),
|
|
|
|
|
):
|
|
|
|
|
lines.append(f" <!-- {label or '?'} -->")
|
|
|
|
|
if decl is None:
|
|
|
|
|
lines += [
|
|
|
|
|
" (the module no longer declares this template)",
|
|
|
|
|
"",
|
|
|
|
|
]
|
|
|
|
|
continue
|
|
|
|
|
path, snippet = decl
|
|
|
|
|
lines.append(f" <!-- {path} -->")
|
|
|
|
|
lines += [f" {line}" for line in snippet.splitlines()]
|
|
|
|
|
shape = (
|
|
|
|
|
"inheritance specs (needs inherit_id)"
|
|
|
|
|
if re.search(r"inherit_id\s*=", snippet)
|
|
|
|
|
else "a standalone template"
|
|
|
|
|
)
|
|
|
|
|
lines += [f" -> {shape}", ""]
|
|
|
|
|
lines += [
|
[IMP] migration: say why Odoo changing its own template is our problem
The declarations view showed a 12.0 template becoming a 13.0 extension and
stopped there. Read alone it says « Odoo changed its code », and the obvious
reaction is to ask why that is anyone's problem — which is exactly what it
prompted.
It is not the problem. On a database without a copy the module upgrade
rewrites the view and nothing breaks. It breaks because a COPY exists and
froze the old shape. That sentence was missing, and so was the other half of
the answer: what the copy actually holds. A copy identical to the view it
shadows costs nothing to neutralize; one carrying five lines of theme hooks
costs those five lines. Both are now stated where the question arises.
Checked on the database mid-migration: the copy that raised the question
reports +5/-3, and the two other cases — identical to its twin, and a page
with no twin at all — each say so.
--- FR ---
La vue des déclarations montrait un gabarit 12.0 devenant une extension 13.0
et s'arrêtait là. Lue seule, elle dit « Odoo a changé son code », et la
réaction naturelle est de demander en quoi cela nous regarde — ce qu'elle a
justement provoqué.
Ce n'est pas le problème. Sur une base sans copie, la mise à jour du module
réécrit la vue et rien ne casse. Ça casse parce qu'une COPIE existe et a figé
l'ancienne forme. Cette phrase manquait, et l'autre moitié de la réponse
aussi : ce que la copie contient réellement. Une copie identique à la vue
qu'elle masque ne coûte rien à neutraliser ; une qui porte cinq lignes
d'ancres de thème coûte ces cinq lignes. Les deux sont dites là où la question
se pose.
Vérifié sur la base en cours de migration : la copie qui a soulevé la question
rapporte +5/-3, et les deux autres cas — identique à sa jumelle, et page sans
jumelle — le disent chacun.
Assisted-by: Claude Opus 5
2026-08-12 01:28:50 -04:00
|
|
|
" Odoo changing the shape of its own template is NOT the problem: on a",
|
|
|
|
|
" database without a copy, the module upgrade rewrites the view and",
|
|
|
|
|
" nothing breaks. It breaks here because a COPY exists and froze the",
|
|
|
|
|
" old shape — the copy follows the module and becomes an extension,",
|
|
|
|
|
" while still holding a standalone template. Odoo then applies that",
|
|
|
|
|
" template as an inheritance spec and stops on « cannot be located in",
|
|
|
|
|
" parent view ».",
|
|
|
|
|
"",
|
[ADD] migration: look at a COW copy before agreeing to give it up
The prompt asked whether to neutralize copies without showing what they hold.
Answering meant giving up a customization sight unseen — often three lines, an
id and a container width, sometimes a whole page, and nothing told them apart.
« v » now shows both halves of the question. What the copy changed, diffed
against the module view it shadows: on the database at hand, 5 lines added and
3 removed, two CSS anchors and a container width. And why it breaks, by
showing the declaration in each version — portal.frontend_layout is a
standalone template in 12.0 and inheritance specs in 13.0, which is the whole
explanation. « w » is the same two, full screen, space to switch.
The current version comes from ir_module_module, not .odoo-version: the
checkout is switched to the target before this runs, so reading the file
compared the target with itself and printed the same declaration twice. That
is what it did until it was run against a real migration.
--list says which copies a past neutralization put aside. A renamed key is
invisible in the interface, so without it the only trace was remembering.
Checked on the VM mid-migration: both views on view 2670, the screen builds
headless and toggles, --list reports and reports nothing when there is
nothing. A missing database now exits 2 with a message instead of a traceback.
--- FR ---
L'invite demandait de neutraliser des copies sans montrer ce qu'elles
contiennent. Répondre revenait à renoncer à une personnalisation sans l'avoir
vue — souvent trois lignes, un id et une largeur de conteneur, parfois une
page entière, et rien ne les distinguait.
« v » montre désormais les deux moitiés de la question. Ce que la copie a
changé, comparé à la vue de module qu'elle masque : sur la base en cours, 5
lignes ajoutées et 3 retirées, deux ancres CSS et une largeur. Et pourquoi ça
casse, en affichant la déclaration dans chaque version —
portal.frontend_layout est un gabarit autonome en 12.0 et des consignes
d'héritage en 13.0, ce qui est toute l'explication. « w » donne les deux en
plein écran, espace pour basculer.
La version courante vient d'ir_module_module, pas de .odoo-version : le
checkout est basculé sur la cible avant cette étape, donc lire le fichier
comparait la cible avec elle-même et affichait deux fois la même déclaration.
C'est ce qu'il faisait jusqu'à l'essai sur une vraie migration.
--list dit quelles copies une neutralisation passée a mises de côté. Une clé
renommée est invisible dans l'interface ; sans cela, la seule trace était de
s'en souvenir.
Vérifié sur la VM en cours de migration : les deux vues sur la vue 2670,
l'écran se construit sans terminal et bascule, --list rapporte et ne rapporte
rien quand il n'y a rien. Une base absente sort en 2 avec un message plutôt
qu'une trace d'appel.
Assisted-by: Claude Opus 5
2026-08-11 23:49:05 -04:00
|
|
|
]
|
[IMP] migration: say why Odoo changing its own template is our problem
The declarations view showed a 12.0 template becoming a 13.0 extension and
stopped there. Read alone it says « Odoo changed its code », and the obvious
reaction is to ask why that is anyone's problem — which is exactly what it
prompted.
It is not the problem. On a database without a copy the module upgrade
rewrites the view and nothing breaks. It breaks because a COPY exists and
froze the old shape. That sentence was missing, and so was the other half of
the answer: what the copy actually holds. A copy identical to the view it
shadows costs nothing to neutralize; one carrying five lines of theme hooks
costs those five lines. Both are now stated where the question arises.
Checked on the database mid-migration: the copy that raised the question
reports +5/-3, and the two other cases — identical to its twin, and a page
with no twin at all — each say so.
--- FR ---
La vue des déclarations montrait un gabarit 12.0 devenant une extension 13.0
et s'arrêtait là. Lue seule, elle dit « Odoo a changé son code », et la
réaction naturelle est de demander en quoi cela nous regarde — ce qu'elle a
justement provoqué.
Ce n'est pas le problème. Sur une base sans copie, la mise à jour du module
réécrit la vue et rien ne casse. Ça casse parce qu'une COPIE existe et a figé
l'ancienne forme. Cette phrase manquait, et l'autre moitié de la réponse
aussi : ce que la copie contient réellement. Une copie identique à la vue
qu'elle masque ne coûte rien à neutraliser ; une qui porte cinq lignes
d'ancres de thème coûte ces cinq lignes. Les deux sont dites là où la question
se pose.
Vérifié sur la base en cours de migration : la copie qui a soulevé la question
rapporte +5/-3, et les deux autres cas — identique à sa jumelle, et page sans
jumelle — le disent chacun.
Assisted-by: Claude Opus 5
2026-08-12 01:28:50 -04:00
|
|
|
# Sans ceci, cette vue se lit comme « du code Odoo qui change », et l'on
|
|
|
|
|
# se demande pourquoi c'est notre affaire. Ce qui la rend nôtre est ce que
|
|
|
|
|
# la copie contient — la seule chose que neutraliser ferait perdre.
|
|
|
|
|
lines += _what_the_copy_holds(finding)
|
[ADD] migration: look at a COW copy before agreeing to give it up
The prompt asked whether to neutralize copies without showing what they hold.
Answering meant giving up a customization sight unseen — often three lines, an
id and a container width, sometimes a whole page, and nothing told them apart.
« v » now shows both halves of the question. What the copy changed, diffed
against the module view it shadows: on the database at hand, 5 lines added and
3 removed, two CSS anchors and a container width. And why it breaks, by
showing the declaration in each version — portal.frontend_layout is a
standalone template in 12.0 and inheritance specs in 13.0, which is the whole
explanation. « w » is the same two, full screen, space to switch.
The current version comes from ir_module_module, not .odoo-version: the
checkout is switched to the target before this runs, so reading the file
compared the target with itself and printed the same declaration twice. That
is what it did until it was run against a real migration.
--list says which copies a past neutralization put aside. A renamed key is
invisible in the interface, so without it the only trace was remembering.
Checked on the VM mid-migration: both views on view 2670, the screen builds
headless and toggles, --list reports and reports nothing when there is
nothing. A missing database now exits 2 with a message instead of a traceback.
--- FR ---
L'invite demandait de neutraliser des copies sans montrer ce qu'elles
contiennent. Répondre revenait à renoncer à une personnalisation sans l'avoir
vue — souvent trois lignes, un id et une largeur de conteneur, parfois une
page entière, et rien ne les distinguait.
« v » montre désormais les deux moitiés de la question. Ce que la copie a
changé, comparé à la vue de module qu'elle masque : sur la base en cours, 5
lignes ajoutées et 3 retirées, deux ancres CSS et une largeur. Et pourquoi ça
casse, en affichant la déclaration dans chaque version —
portal.frontend_layout est un gabarit autonome en 12.0 et des consignes
d'héritage en 13.0, ce qui est toute l'explication. « w » donne les deux en
plein écran, espace pour basculer.
La version courante vient d'ir_module_module, pas de .odoo-version : le
checkout est basculé sur la cible avant cette étape, donc lire le fichier
comparait la cible avec elle-même et affichait deux fois la même déclaration.
C'est ce qu'il faisait jusqu'à l'essai sur une vraie migration.
--list dit quelles copies une neutralisation passée a mises de côté. Une clé
renommée est invisible dans l'interface ; sans cela, la seule trace était de
s'en souvenir.
Vérifié sur la VM en cours de migration : les deux vues sur la vue 2670,
l'écran se construit sans terminal et bascule, --list rapporte et ne rapporte
rien quand il n'y a rien. Une base absente sort en 2 avec un message plutôt
qu'une trace d'appel.
Assisted-by: Claude Opus 5
2026-08-11 23:49:05 -04:00
|
|
|
return "\n".join(lines)
|
|
|
|
|
|
|
|
|
|
|
[IMP] migration: say why Odoo changing its own template is our problem
The declarations view showed a 12.0 template becoming a 13.0 extension and
stopped there. Read alone it says « Odoo changed its code », and the obvious
reaction is to ask why that is anyone's problem — which is exactly what it
prompted.
It is not the problem. On a database without a copy the module upgrade
rewrites the view and nothing breaks. It breaks because a COPY exists and
froze the old shape. That sentence was missing, and so was the other half of
the answer: what the copy actually holds. A copy identical to the view it
shadows costs nothing to neutralize; one carrying five lines of theme hooks
costs those five lines. Both are now stated where the question arises.
Checked on the database mid-migration: the copy that raised the question
reports +5/-3, and the two other cases — identical to its twin, and a page
with no twin at all — each say so.
--- FR ---
La vue des déclarations montrait un gabarit 12.0 devenant une extension 13.0
et s'arrêtait là. Lue seule, elle dit « Odoo a changé son code », et la
réaction naturelle est de demander en quoi cela nous regarde — ce qu'elle a
justement provoqué.
Ce n'est pas le problème. Sur une base sans copie, la mise à jour du module
réécrit la vue et rien ne casse. Ça casse parce qu'une COPIE existe et a figé
l'ancienne forme. Cette phrase manquait, et l'autre moitié de la réponse
aussi : ce que la copie contient réellement. Une copie identique à la vue
qu'elle masque ne coûte rien à neutraliser ; une qui porte cinq lignes
d'ancres de thème coûte ces cinq lignes. Les deux sont dites là où la question
se pose.
Vérifié sur la base en cours de migration : la copie qui a soulevé la question
rapporte +5/-3, et les deux autres cas — identique à sa jumelle, et page sans
jumelle — le disent chacun.
Assisted-by: Claude Opus 5
2026-08-12 01:28:50 -04:00
|
|
|
def _what_the_copy_holds(finding):
|
|
|
|
|
"""Ce que la copie porte en propre, dit en une ou deux lignes.
|
|
|
|
|
|
|
|
|
|
Relie les deux moitiés de la question : le mécanisme explique POURQUOI ça
|
|
|
|
|
casse, ceci dit CE QU'IL EN COÛTE. Une copie identique à sa jumelle ne
|
|
|
|
|
coûte rien, et le savoir change la décision.
|
|
|
|
|
"""
|
|
|
|
|
if finding["module_id"] is None:
|
|
|
|
|
return [
|
|
|
|
|
" This copy has no module view of that name: it is a page made in",
|
|
|
|
|
" the website editor, and nothing else holds its content.",
|
|
|
|
|
]
|
|
|
|
|
left = finding["module_arch"].splitlines()
|
|
|
|
|
right = finding["copy_arch"].splitlines()
|
|
|
|
|
diff = [
|
|
|
|
|
line
|
|
|
|
|
for line in difflib.unified_diff(left, right, lineterm="", n=0)
|
|
|
|
|
if line[:1] in "+-" and not line.startswith(("+++", "---"))
|
|
|
|
|
]
|
|
|
|
|
if not diff:
|
|
|
|
|
return [
|
|
|
|
|
" This copy is IDENTICAL to the module view it shadows: it holds no",
|
|
|
|
|
" customization at all, so neutralizing it loses nothing.",
|
|
|
|
|
]
|
|
|
|
|
n_plus = sum(1 for x in diff if x.startswith("+"))
|
|
|
|
|
n_minus = sum(1 for x in diff if x.startswith("-"))
|
|
|
|
|
return [
|
|
|
|
|
f" This copy differs from the module view by +{n_plus}/-{n_minus}"
|
|
|
|
|
" line(s):",
|
|
|
|
|
" that is the customization, and all that neutralizing gives up.",
|
|
|
|
|
" Run without --shape to read it.",
|
|
|
|
|
]
|
|
|
|
|
|
|
|
|
|
|
[ADD] migration: look at a COW copy before agreeing to give it up
The prompt asked whether to neutralize copies without showing what they hold.
Answering meant giving up a customization sight unseen — often three lines, an
id and a container width, sometimes a whole page, and nothing told them apart.
« v » now shows both halves of the question. What the copy changed, diffed
against the module view it shadows: on the database at hand, 5 lines added and
3 removed, two CSS anchors and a container width. And why it breaks, by
showing the declaration in each version — portal.frontend_layout is a
standalone template in 12.0 and inheritance specs in 13.0, which is the whole
explanation. « w » is the same two, full screen, space to switch.
The current version comes from ir_module_module, not .odoo-version: the
checkout is switched to the target before this runs, so reading the file
compared the target with itself and printed the same declaration twice. That
is what it did until it was run against a real migration.
--list says which copies a past neutralization put aside. A renamed key is
invisible in the interface, so without it the only trace was remembering.
Checked on the VM mid-migration: both views on view 2670, the screen builds
headless and toggles, --list reports and reports nothing when there is
nothing. A missing database now exits 2 with a message instead of a traceback.
--- FR ---
L'invite demandait de neutraliser des copies sans montrer ce qu'elles
contiennent. Répondre revenait à renoncer à une personnalisation sans l'avoir
vue — souvent trois lignes, un id et une largeur de conteneur, parfois une
page entière, et rien ne les distinguait.
« v » montre désormais les deux moitiés de la question. Ce que la copie a
changé, comparé à la vue de module qu'elle masque : sur la base en cours, 5
lignes ajoutées et 3 retirées, deux ancres CSS et une largeur. Et pourquoi ça
casse, en affichant la déclaration dans chaque version —
portal.frontend_layout est un gabarit autonome en 12.0 et des consignes
d'héritage en 13.0, ce qui est toute l'explication. « w » donne les deux en
plein écran, espace pour basculer.
La version courante vient d'ir_module_module, pas de .odoo-version : le
checkout est basculé sur la cible avant cette étape, donc lire le fichier
comparait la cible avec elle-même et affichait deux fois la même déclaration.
C'est ce qu'il faisait jusqu'à l'essai sur une vraie migration.
--list dit quelles copies une neutralisation passée a mises de côté. Une clé
renommée est invisible dans l'interface ; sans cela, la seule trace était de
s'en souvenir.
Vérifié sur la VM en cours de migration : les deux vues sur la vue 2670,
l'écran se construit sans terminal et bascule, --list rapporte et ne rapporte
rien quand il n'y a rien. Une base absente sort en 2 avec un message plutôt
qu'une trace d'appel.
Assisted-by: Claude Opus 5
2026-08-11 23:49:05 -04:00
|
|
|
def render_all(lst_finding, shape=False):
|
|
|
|
|
"""The whole report, one block per finding."""
|
|
|
|
|
if not lst_finding:
|
|
|
|
|
return "✅ No website COW view is at risk.\n"
|
|
|
|
|
render = render_shape if shape else render_diff
|
|
|
|
|
return "\n\n".join(render(f) for f in lst_finding) + "\n"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def main(argv=None):
|
|
|
|
|
parser = argparse.ArgumentParser(
|
|
|
|
|
description=(
|
|
|
|
|
"Show what each at-risk website COW copy holds, and why the next"
|
|
|
|
|
" version breaks it (read-only)."
|
|
|
|
|
)
|
|
|
|
|
)
|
|
|
|
|
parser.add_argument("-d", "--database", required=True)
|
|
|
|
|
parser.add_argument(
|
|
|
|
|
"-t",
|
|
|
|
|
"--target_version",
|
|
|
|
|
required=True,
|
|
|
|
|
help="target Odoo source directory, e.g. odoo13.0",
|
|
|
|
|
)
|
|
|
|
|
parser.add_argument(
|
|
|
|
|
"--current",
|
|
|
|
|
default=None,
|
|
|
|
|
help="current Odoo source directory (default: from .odoo-version)",
|
|
|
|
|
)
|
|
|
|
|
parser.add_argument(
|
|
|
|
|
"--shape",
|
|
|
|
|
action="store_true",
|
|
|
|
|
help="show the declarations instead of the customization diff",
|
|
|
|
|
)
|
|
|
|
|
parser.add_argument(
|
|
|
|
|
"--tui",
|
|
|
|
|
action="store_true",
|
|
|
|
|
help="browse full screen, switching between the two views",
|
|
|
|
|
)
|
|
|
|
|
config = parser.parse_args(argv)
|
|
|
|
|
|
|
|
|
|
if not os.path.isdir(config.target_version):
|
|
|
|
|
print(f"❌ Target version '{config.target_version}' not found.")
|
|
|
|
|
return 2
|
|
|
|
|
try:
|
|
|
|
|
lst_finding = collect(
|
|
|
|
|
config.database, config.target_version, config.current
|
|
|
|
|
)
|
|
|
|
|
except RuntimeError as exc:
|
|
|
|
|
print(f"❌ {exc}")
|
|
|
|
|
return 2
|
|
|
|
|
|
|
|
|
|
if config.tui and lst_finding:
|
|
|
|
|
from cow_drift_tui import run_tui
|
|
|
|
|
|
|
|
|
|
if run_tui(lst_finding):
|
|
|
|
|
return 1
|
|
|
|
|
print(render_all(lst_finding, shape=config.shape))
|
|
|
|
|
return 1 if lst_finding else 0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
if __name__ == "__main__":
|
|
|
|
|
sys.exit(main())
|