summaryrefslogtreecommitdiffstats
path: root/bin/reproducible_html_live_status.py
blob: 78187be5404c8cefa68b818eb2620ecdf38ef313 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
#!/usr/bin/python3
# -*- coding: utf-8 -*-
#
# Copyright © 2015 Holger Levsen <holger@layer-acht.org>
# based on ~jenkins.d.n:~mattia/status.sh by Mattia Rizzolo <mattia@mapreri.org>
# Licensed under GPL-2
#
# Depends: python3
#

from reproducible_common import *
from reproducible_html_indexes import build_leading_text_section
import glob

def generate_schedule(arch):
    """ the schedule pages are very different than others index pages """
    log.info('Building the schedule index page for ' + arch + '...')
    title = 'Packages currently scheduled on ' + arch + ' for testing for build reproducibility'
    query = 'SELECT sch.date_scheduled, s.suite, s.architecture, s.name, ' + \
            'r.status, r.build_duration, ' + \
            '(SELECT coalesce(AVG(h.build_duration), 0) FROM stats_build AS h WHERE h.status IN ("reproducible", "unreproducible") AND h.name=s.name AND h.suite=s.suite AND h.architecture=s.architecture) ' + \
            'FROM schedule AS sch JOIN sources AS s ON sch.package_id=s.id LEFT JOIN results AS r ON s.id=r.package_id ' + \
            'WHERE sch.date_build_started IS NULL AND s.architecture="{arch}" ORDER BY sch.date_scheduled'
    # 'AND h.name=s.name AND h.suite=s.suite AND h.architecture=s.architecture' in this query and the query below is needed due to not using package_id in the stats_build table, which should be fixed...
    text = Template('$tot packages are currently scheduled for testing on $arch:')
    html = ''
    rows = query_db(query.format(arch=arch))
    html += build_leading_text_section({'text': text}, rows, defaultsuite, arch)
    html += generate_live_status_table(arch)
    html += '<p><table class="scheduled">\n' + tab
    html += '<tr><th>#</th><th>scheduled at</th><th>suite</th>'
    html += '<th>arch</th><th>source package</th><th>previous build status</th><th>previous build duration</th><th>average build duration</th></tr>\n'
    bugs = get_bugs()
    for row in rows:
        # 0: date_scheduled, 1: suite, 2: arch, 3: pkg name 4: previous status 5: previous build duration 6. avg build duration
        pkg = row[3]
        duration = convert_into_hms_string(row[5])
        avg_duration = convert_into_hms_string(row[6])
        html += tab + '<tr><td>&nbsp;</td><td>' + row[0] + '</td>'
        html += '<td>' + row[1] + '</td><td>' + row[2] + '</td><td><code>'
        html += link_package(pkg, row[1], row[2], bugs)
        html += '</code></td><td>'+str(row[4])+'</td><td>'+duration+'</td><td>' + avg_duration + '</td></tr>\n'
    html += '</table></p>\n'
    destfile = BASE + '/index_' + arch + '_scheduled.html'
    desturl = REPRODUCIBLE_URL + '/index_' + arch + '_scheduled.html'
    write_html_page(title=title, body=html, destfile=destfile, arch=arch, style_note=True, refresh_every=60)
    log.info("Page generated at " + desturl)


def generate_live_status_table(arch):
    query = 'SELECT s.id, s.suite, s.architecture, s.name, s.version, ' + \
            'p.date_build_started, r.status, r.build_duration, ' + \
            '(SELECT coalesce(AVG(h.build_duration), 0) FROM stats_build AS h WHERE h.status IN ("reproducible", "unreproducible") AND h.name=s.name AND h.suite=s.suite AND h.architecture=s.architecture) ' + \
            ', p.job ' + \
            'FROM sources AS s JOIN schedule AS p ON p.package_id=s.id LEFT JOIN results AS r ON s.id=r.package_id ' + \
            'WHERE p.date_build_started IS NOT NULL AND s.architecture="{arch}" ' + \
            'ORDER BY p.date_build_started DESC'
    html = ''
    rows = query_db(query.format(arch=arch))
    html += '<p><table class="scheduled">\n' + tab
    html += '<tr><th>#</th><th>src pkg id</th><th>suite</th><th>arch</th>'
    html += '<th>source package</th><th>version</th></th>'
    html += '<th>build started</th><th>previous build status</th>'
    html += '<th>previous build duration</th><th>average build duration</th><th>builder job</th>'
    html += '</tr>\n'
    counter = 0
    # the path should probably not be hard coded here…
    builders = len(glob.glob('/var/lib/jenkins/jobs/reproducible_builder_' + arch + '_*'))
    for row in rows:
        counter += 1
        if counter > builders:
             html += '<tr><td colspan="10">There are more builds marked as currently building in the database (' + str(counter) + ') than there are ' + arch + ' build jobs (' + str(builders) + '). This does not compute, please investigate and fix the cause.</td></tr>'
        elif builders == 0:
             html += '<tr><td colspan="10">0 build jobs for ' + arch + ' detected. This does not compute, please investigate and fix the cause.</td></tr>'
        suite = row[1]
        arch = row[2]
        pkg = row[3]
        duration = convert_into_hms_string(row[7])
        avg_duration = convert_into_hms_string(row[8])
        html += tab + '<tr><td>&nbsp;</td><td>' + str(row[0]) + '</td>'
        html += '<td>' + suite + '</td><td>' + arch + '</td>'
        html += '<td><code>' + link_package(pkg, suite, arch) + '</code></td>'
        html += '<td>' + str(row[4]) + '</td><td>' + str(row[5]) + '</td>'
        html += '<td>' + str(row[6]) + '</td><td>' + duration + '</td><td>' + avg_duration + '</td>'
        html += '<td><a href="https://jenkins.debian.net/job/reproducible_builder_' + str(row[9]) + '/console">' + str(row[9]) + '</a></td>'
        html += '</tr>\n'
    html += '</table></p>\n'
    return html

if __name__ == '__main__':
    for arch in ARCHS:
        generate_schedule(arch)