blob: 865390a6f351d7c279c1d9b4cfb5358c2935fb88 [file] [log] [blame]
Simon Glassfc3fe1c2013-04-03 11:07:16 +00001# Copyright (c) 2013 The Chromium OS Authors.
2#
Wolfgang Denk1a459662013-07-08 09:37:19 +02003# SPDX-License-Identifier: GPL-2.0+
Simon Glassfc3fe1c2013-04-03 11:07:16 +00004#
5
Simon Glass6eede342014-08-09 15:32:58 -06006(Please read 'How to change from MAKEALL' if you are used to that tool)
7
Simon Glassfc3fe1c2013-04-03 11:07:16 +00008What is this?
9=============
10
11This tool handles building U-Boot to check that you have not broken it
12with your patch series. It can build each individual commit and report
13which boards fail on which commits, and which errors come up. It aims
14to make full use of multi-processor machines.
15
16A key feature of buildman is its output summary, which allows warnings,
17errors or image size increases in a particular commit or board to be
18quickly identified and the offending commit pinpointed. This can be a big
19help for anyone working with >10 patches at a time.
20
21
22Caveats
23=======
24
25Buildman is still in its infancy. It is already a very useful tool, but
26expect to find problems and send patches.
27
28Buildman can be stopped and restarted, in which case it will continue
29where it left off. This should happen cleanly and without side-effects.
30If not, it is a bug, for which a patch would be welcome.
31
32Buildman gets so tied up in its work that it can ignore the outside world.
33You may need to press Ctrl-C several times to quit it. Also it will print
34out various exceptions when stopped.
35
36
37Theory of Operation
38===================
39
40(please read this section in full twice or you will be perpetually confused)
41
42Buildman is a builder. It is not make, although it runs make. It does not
43produce any useful output on the terminal while building, except for
Simon Glasse5a0e5d2014-08-09 15:33:03 -060044progress information (except with -v, see below). All the output (errors,
Dirk Behme3e1ded12014-12-23 07:41:26 +010045warnings and binaries if you ask for them) is stored in output
Simon Glasse5a0e5d2014-08-09 15:33:03 -060046directories, which you can look at while the build is progressing, or when
47it is finished.
Simon Glassfc3fe1c2013-04-03 11:07:16 +000048
49Buildman produces a concise summary of which boards succeeded and failed.
50It shows which commit introduced which board failure using a simple
51red/green colour coding. Full error information can be requested, in which
52case it is de-duped and displayed against the commit that introduced the
53error. An example workflow is below.
54
55Buildman stores image size information and can report changes in image size
56from commit to commit. An example of this is below.
57
58Buildman starts multiple threads, and each thread builds for one board at
59a time. A thread starts at the first commit, configures the source for your
60board and builds it. Then it checks out the next commit and does an
61incremental build. Eventually the thread reaches the last commit and stops.
62If errors or warnings are found along the way, the thread will reconfigure
63after every commit, and your build will be very slow. This is because a
64file that produces just a warning would not normally be rebuilt in an
65incremental build.
66
67Buildman works in an entirely separate place from your U-Boot repository.
68It creates a separate working directory for each thread, and puts the
69output files in the working directory, organised by commit name and board
70name, in a two-level hierarchy.
71
72Buildman is invoked in your U-Boot directory, the one with the .git
73directory. It clones this repository into a copy for each thread, and the
74threads do not affect the state of your git repository. Any checkouts done
75by the thread affect only the working directory for that thread.
76
Simon Glasscec83c32014-08-09 15:32:57 -060077Buildman automatically selects the correct tool chain for each board. You
78must supply suitable tool chains, but buildman takes care of selecting the
Simon Glassfc3fe1c2013-04-03 11:07:16 +000079right one.
80
Simon Glasse5a0e5d2014-08-09 15:33:03 -060081Buildman generally builds a branch (with the -b flag), and in this case
82builds the upstream commit as well, for comparison. It cannot build
83individual commits at present, unless (maybe) you point it at an empty
84branch. Put all your commits in a branch, set the branch's upstream to a
85valid value, and all will be well. Otherwise buildman will perform random
86actions. Use -n to check what the random actions might be.
87
Simon Glass1d8104f2014-10-16 01:05:56 -060088If you just want to build the current source tree, leave off the -b flag
89and add -e. This will display results and errors as they happen. You can
90still look at them later using -se. Note that buildman will assume that the
91source has changed, and will build all specified boards in this case.
Simon Glassfc3fe1c2013-04-03 11:07:16 +000092
93Buildman is optimised for building many commits at once, for many boards.
94On multi-core machines, Buildman is fast because it uses most of the
95available CPU power. When it gets to the end, or if you are building just
96a few commits or boards, it will be pretty slow. As a tip, if you don't
97plan to use your machine for anything else, you can use -T to increase the
98number of threads beyond the default.
99
Stephen Warren8426d8b2013-10-10 10:00:20 -0600100Buildman lets you build all boards, or a subset. Specify the subset by passing
101command-line arguments that list the desired board name, architecture name,
102SOC name, or anything else in the boards.cfg file. Multiple arguments are
103allowed. Each argument will be interpreted as a regular expression, so
104behaviour is a superset of exact or substring matching. Examples are:
105
106* 'tegra20' All boards with a Tegra20 SoC
107* 'tegra' All boards with any Tegra Soc (Tegra20, Tegra30, Tegra114...)
108* '^tegra[23]0$' All boards with either Tegra20 or Tegra30 SoC
109* 'powerpc' All PowerPC boards
Simon Glassfc3fe1c2013-04-03 11:07:16 +0000110
Simon Glass6131bea2014-08-09 15:33:08 -0600111While the default is to OR the terms together, you can also make use of
112the '&' operator to limit the selection:
113
114* 'freescale & arm sandbox' All Freescale boards with ARM architecture,
115 plus sandbox
116
Simon Glass3cf4ae62014-08-28 09:43:41 -0600117You can also use -x to specifically exclude some boards. For example:
118
119 buildmand arm -x nvidia,freescale,.*ball$
120
121means to build all arm boards except nvidia, freescale and anything ending
122with 'ball'.
123
Dirk Behme3e1ded12014-12-23 07:41:26 +0100124It is convenient to use the -n option to see what will be built based on
Simon Glass6131bea2014-08-09 15:33:08 -0600125the subset given.
126
Simon Glassfc3fe1c2013-04-03 11:07:16 +0000127Buildman does not store intermediate object files. It optionally copies
128the binary output into a directory when a build is successful. Size
129information is always recorded. It needs a fair bit of disk space to work,
130typically 250MB per thread.
131
132
133Setting up
134==========
135
1361. Get the U-Boot source. You probably already have it, but if not these
137steps should get you started with a repo and some commits for testing.
138
139$ cd /path/to/u-boot
140$ git clone git://git.denx.de/u-boot.git .
141$ git checkout -b my-branch origin/master
142$ # Add some commits to the branch, reading for testing
143
Simon Glass62005342014-12-01 17:34:02 -07001442. Create ~/.buildman to tell buildman where to find tool chains (see 'The
145.buildman file' later for details). As an example:
Simon Glassfc3fe1c2013-04-03 11:07:16 +0000146
147# Buildman settings file
148
149[toolchain]
150root: /
151rest: /toolchains/*
152eldk: /opt/eldk-4.2
Simon Glasse9569472014-08-09 15:33:07 -0600153arm: /opt/linaro/gcc-linaro-arm-linux-gnueabihf-4.8-2013.08_linux
154aarch64: /opt/linaro/gcc-linaro-aarch64-none-elf-4.8-2013.10_linux
Simon Glassfc3fe1c2013-04-03 11:07:16 +0000155
156[toolchain-alias]
157x86: i386
158blackfin: bfin
159sh: sh4
160nds32: nds32le
161openrisc: or32
162
163
164This selects the available toolchain paths. Add the base directory for
165each of your toolchains here. Buildman will search inside these directories
166and also in any '/usr' and '/usr/bin' subdirectories.
167
168Make sure the tags (here root: rest: and eldk:) are unique.
169
170The toolchain-alias section indicates that the i386 toolchain should be used
171to build x86 commits.
172
173
Simon Glass34699692014-12-01 17:34:01 -07001743. Make sure you have the require Python pre-requisites
175
176Buildman uses multiprocessing, Queue, shutil, StringIO and ConfigParser.
177These should normally be available, but if you get an error like this then
178you will need to obtain those modules:
179
180 ImportError: No module named multiprocessing
181
182
1834. Check the available toolchains
Simon Glassfc3fe1c2013-04-03 11:07:16 +0000184
185Run this check to make sure that you have a toolchain for every architecture.
186
187$ ./tools/buildman/buildman --list-tool-chains
188Scanning for tool chains
189 - scanning path '/'
190 - looking in '/.'
191 - looking in '/bin'
192 - looking in '/usr/bin'
193 - found '/usr/bin/gcc'
194Tool chain test: OK
195 - found '/usr/bin/c89-gcc'
196Tool chain test: OK
197 - found '/usr/bin/c99-gcc'
198Tool chain test: OK
199 - found '/usr/bin/x86_64-linux-gnu-gcc'
200Tool chain test: OK
201 - scanning path '/toolchains/powerpc-linux'
202 - looking in '/toolchains/powerpc-linux/.'
203 - looking in '/toolchains/powerpc-linux/bin'
204 - found '/toolchains/powerpc-linux/bin/powerpc-linux-gcc'
205Tool chain test: OK
206 - looking in '/toolchains/powerpc-linux/usr/bin'
207 - scanning path '/toolchains/nds32le-linux-glibc-v1f'
208 - looking in '/toolchains/nds32le-linux-glibc-v1f/.'
209 - looking in '/toolchains/nds32le-linux-glibc-v1f/bin'
210 - found '/toolchains/nds32le-linux-glibc-v1f/bin/nds32le-linux-gcc'
211Tool chain test: OK
212 - looking in '/toolchains/nds32le-linux-glibc-v1f/usr/bin'
213 - scanning path '/toolchains/nios2'
214 - looking in '/toolchains/nios2/.'
215 - looking in '/toolchains/nios2/bin'
216 - found '/toolchains/nios2/bin/nios2-linux-gcc'
217Tool chain test: OK
218 - found '/toolchains/nios2/bin/nios2-linux-uclibc-gcc'
219Tool chain test: OK
220 - looking in '/toolchains/nios2/usr/bin'
221 - found '/toolchains/nios2/usr/bin/nios2-linux-gcc'
222Tool chain test: OK
223 - found '/toolchains/nios2/usr/bin/nios2-linux-uclibc-gcc'
224Tool chain test: OK
225 - scanning path '/toolchains/microblaze-unknown-linux-gnu'
226 - looking in '/toolchains/microblaze-unknown-linux-gnu/.'
227 - looking in '/toolchains/microblaze-unknown-linux-gnu/bin'
228 - found '/toolchains/microblaze-unknown-linux-gnu/bin/microblaze-unknown-linux-gnu-gcc'
229Tool chain test: OK
230 - found '/toolchains/microblaze-unknown-linux-gnu/bin/mb-linux-gcc'
231Tool chain test: OK
232 - looking in '/toolchains/microblaze-unknown-linux-gnu/usr/bin'
233 - scanning path '/toolchains/mips-linux'
234 - looking in '/toolchains/mips-linux/.'
235 - looking in '/toolchains/mips-linux/bin'
236 - found '/toolchains/mips-linux/bin/mips-linux-gcc'
237Tool chain test: OK
238 - looking in '/toolchains/mips-linux/usr/bin'
239 - scanning path '/toolchains/old'
240 - looking in '/toolchains/old/.'
241 - looking in '/toolchains/old/bin'
242 - looking in '/toolchains/old/usr/bin'
243 - scanning path '/toolchains/i386-linux'
244 - looking in '/toolchains/i386-linux/.'
245 - looking in '/toolchains/i386-linux/bin'
246 - found '/toolchains/i386-linux/bin/i386-linux-gcc'
247Tool chain test: OK
248 - looking in '/toolchains/i386-linux/usr/bin'
249 - scanning path '/toolchains/bfin-uclinux'
250 - looking in '/toolchains/bfin-uclinux/.'
251 - looking in '/toolchains/bfin-uclinux/bin'
252 - found '/toolchains/bfin-uclinux/bin/bfin-uclinux-gcc'
253Tool chain test: OK
254 - looking in '/toolchains/bfin-uclinux/usr/bin'
255 - scanning path '/toolchains/sparc-elf'
256 - looking in '/toolchains/sparc-elf/.'
257 - looking in '/toolchains/sparc-elf/bin'
258 - found '/toolchains/sparc-elf/bin/sparc-elf-gcc'
259Tool chain test: OK
260 - looking in '/toolchains/sparc-elf/usr/bin'
261 - scanning path '/toolchains/arm-2010q1'
262 - looking in '/toolchains/arm-2010q1/.'
263 - looking in '/toolchains/arm-2010q1/bin'
264 - found '/toolchains/arm-2010q1/bin/arm-none-linux-gnueabi-gcc'
265Tool chain test: OK
266 - looking in '/toolchains/arm-2010q1/usr/bin'
267 - scanning path '/toolchains/from'
268 - looking in '/toolchains/from/.'
269 - looking in '/toolchains/from/bin'
270 - looking in '/toolchains/from/usr/bin'
271 - scanning path '/toolchains/sh4-gentoo-linux-gnu'
272 - looking in '/toolchains/sh4-gentoo-linux-gnu/.'
273 - looking in '/toolchains/sh4-gentoo-linux-gnu/bin'
274 - found '/toolchains/sh4-gentoo-linux-gnu/bin/sh4-gentoo-linux-gnu-gcc'
275Tool chain test: OK
276 - looking in '/toolchains/sh4-gentoo-linux-gnu/usr/bin'
277 - scanning path '/toolchains/avr32-linux'
278 - looking in '/toolchains/avr32-linux/.'
279 - looking in '/toolchains/avr32-linux/bin'
280 - found '/toolchains/avr32-linux/bin/avr32-gcc'
281Tool chain test: OK
282 - looking in '/toolchains/avr32-linux/usr/bin'
283 - scanning path '/toolchains/m68k-linux'
284 - looking in '/toolchains/m68k-linux/.'
285 - looking in '/toolchains/m68k-linux/bin'
286 - found '/toolchains/m68k-linux/bin/m68k-linux-gcc'
287Tool chain test: OK
288 - looking in '/toolchains/m68k-linux/usr/bin'
289List of available toolchains (17):
290arm : /toolchains/arm-2010q1/bin/arm-none-linux-gnueabi-gcc
291avr32 : /toolchains/avr32-linux/bin/avr32-gcc
292bfin : /toolchains/bfin-uclinux/bin/bfin-uclinux-gcc
293c89 : /usr/bin/c89-gcc
294c99 : /usr/bin/c99-gcc
295i386 : /toolchains/i386-linux/bin/i386-linux-gcc
296m68k : /toolchains/m68k-linux/bin/m68k-linux-gcc
297mb : /toolchains/microblaze-unknown-linux-gnu/bin/mb-linux-gcc
298microblaze: /toolchains/microblaze-unknown-linux-gnu/bin/microblaze-unknown-linux-gnu-gcc
299mips : /toolchains/mips-linux/bin/mips-linux-gcc
300nds32le : /toolchains/nds32le-linux-glibc-v1f/bin/nds32le-linux-gcc
301nios2 : /toolchains/nios2/bin/nios2-linux-gcc
302powerpc : /toolchains/powerpc-linux/bin/powerpc-linux-gcc
303sandbox : /usr/bin/gcc
304sh4 : /toolchains/sh4-gentoo-linux-gnu/bin/sh4-gentoo-linux-gnu-gcc
305sparc : /toolchains/sparc-elf/bin/sparc-elf-gcc
306x86_64 : /usr/bin/x86_64-linux-gnu-gcc
307
308
309You can see that everything is covered, even some strange ones that won't
310be used (c88 and c99). This is a feature.
311
312
313How to run it
314=============
315
316First do a dry run using the -n flag: (replace <branch> with a real, local
317branch with a valid upstream)
318
319$ ./tools/buildman/buildman -b <branch> -n
320
321If it can't detect the upstream branch, try checking out the branch, and
Simon Glass2a9e2c62014-12-01 17:33:54 -0700322doing something like 'git branch --set-upstream-to upstream/master'
323or something similar. Buildman will try to guess a suitable upstream branch
324if it can't find one (you will see a message like" Guessing upstream as ...).
Simon Glassfc3fe1c2013-04-03 11:07:16 +0000325
Simon Glasscec83c32014-08-09 15:32:57 -0600326As an example:
Simon Glassfc3fe1c2013-04-03 11:07:16 +0000327
328Dry run, so not doing much. But I would do this:
329
330Building 18 commits for 1059 boards (4 threads, 1 job per thread)
331Build directory: ../lcd9b
332 5bb3505 Merge branch 'master' of git://git.denx.de/u-boot-arm
333 c18f1b4 tegra: Use const for pinmux_config_pingroup/table()
334 2f043ae tegra: Add display support to funcmux
335 e349900 tegra: fdt: Add pwm binding and node
336 424a5f0 tegra: fdt: Add LCD definitions for Tegra
337 0636ccf tegra: Add support for PWM
338 a994fe7 tegra: Add SOC support for display/lcd
339 fcd7350 tegra: Add LCD driver
340 4d46e9d tegra: Add LCD support to Nvidia boards
341 991bd48 arm: Add control over cachability of memory regions
342 54e8019 lcd: Add CONFIG_LCD_ALIGNMENT to select frame buffer alignment
343 d92aff7 lcd: Add support for flushing LCD fb from dcache after update
344 dbd0677 tegra: Align LCD frame buffer to section boundary
345 0cff9b8 tegra: Support control of cache settings for LCD
346 9c56900 tegra: fdt: Add LCD definitions for Seaboard
347 5cc29db lcd: Add CONFIG_CONSOLE_SCROLL_LINES option to speed console
348 cac5a23 tegra: Enable display/lcd support on Seaboard
349 49ff541 wip
350
351Total boards to build for each commit: 1059
352
353This shows that it will build all 1059 boards, using 4 threads (because
354we have a 4-core CPU). Each thread will run with -j1, meaning that each
355make job will use a single CPU. The list of commits to be built helps you
356confirm that things look about right. Notice that buildman has chosen a
357'base' directory for you, immediately above your source tree.
358
359Buildman works entirely inside the base directory, here ../lcd9b,
360creating a working directory for each thread, and creating output
361directories for each commit and board.
362
363
364Suggested Workflow
365==================
366
367To run the build for real, take off the -n:
368
369$ ./tools/buildman/buildman -b <branch>
370
371Buildman will set up some working directories, and get started. After a
372minute or so it will settle down to a steady pace, with a display like this:
373
374Building 18 commits for 1059 boards (4 threads, 1 job per thread)
375 528 36 124 /19062 1:13:30 : SIMPC8313_SP
376
377This means that it is building 19062 board/commit combinations. So far it
Simon Glasscec83c32014-08-09 15:32:57 -0600378has managed to successfully build 528. Another 36 have built with warnings,
Simon Glassfc3fe1c2013-04-03 11:07:16 +0000379and 124 more didn't build at all. Buildman expects to complete the process
380in an hour and 15 minutes. Use this time to buy a faster computer.
381
382
383To find out how the build went, ask for a summary with -s. You can do this
Dirk Behme3e1ded12014-12-23 07:41:26 +0100384either before the build completes (presumably in another terminal) or
Simon Glassfc3fe1c2013-04-03 11:07:16 +0000385afterwards. Let's work through an example of how this is used:
386
387$ ./tools/buildman/buildman -b lcd9b -s
388...
38901: Merge branch 'master' of git://git.denx.de/u-boot-arm
390 powerpc: + galaxy5200_LOWBOOT
39102: tegra: Use const for pinmux_config_pingroup/table()
39203: tegra: Add display support to funcmux
39304: tegra: fdt: Add pwm binding and node
39405: tegra: fdt: Add LCD definitions for Tegra
39506: tegra: Add support for PWM
39607: tegra: Add SOC support for display/lcd
39708: tegra: Add LCD driver
39809: tegra: Add LCD support to Nvidia boards
39910: arm: Add control over cachability of memory regions
40011: lcd: Add CONFIG_LCD_ALIGNMENT to select frame buffer alignment
40112: lcd: Add support for flushing LCD fb from dcache after update
402 arm: + lubbock
40313: tegra: Align LCD frame buffer to section boundary
40414: tegra: Support control of cache settings for LCD
40515: tegra: fdt: Add LCD definitions for Seaboard
40616: lcd: Add CONFIG_CONSOLE_SCROLL_LINES option to speed console
40717: tegra: Enable display/lcd support on Seaboard
40818: wip
409
410This shows which commits have succeeded and which have failed. In this case
411the build is still in progress so many boards are not built yet (use -u to
412see which ones). But still we can see a few failures. The galaxy5200_LOWBOOT
413never builds correctly. This could be a problem with our toolchain, or it
414could be a bug in the upstream. The good news is that we probably don't need
415to blame our commits. The bad news is it isn't tested on that board.
416
417Commit 12 broke lubbock. That's what the '+ lubbock' means. The failure
418is never fixed by a later commit, or you would see lubbock again, in green,
419without the +.
420
421To see the actual error:
422
423$ ./tools/buildman/buildman -b <branch> -se lubbock
424...
42512: lcd: Add support for flushing LCD fb from dcache after update
426 arm: + lubbock
427+common/libcommon.o: In function `lcd_sync':
428+/u-boot/lcd9b/.bm-work/00/common/lcd.c:120: undefined reference to `flush_dcache_range'
429+arm-none-linux-gnueabi-ld: BFD (Sourcery G++ Lite 2010q1-202) 2.19.51.20090709 assertion fail /scratch/julian/2010q1-release-linux-lite/obj/binutils-src-2010q1-202-arm-none-linux-gnueabi-i686-pc-linux-gnu/bfd/elf32-arm.c:12572
430+make: *** [/u-boot/lcd9b/.bm-work/00/build/u-boot] Error 139
43113: tegra: Align LCD frame buffer to section boundary
43214: tegra: Support control of cache settings for LCD
43315: tegra: fdt: Add LCD definitions for Seaboard
43416: lcd: Add CONFIG_CONSOLE_SCROLL_LINES option to speed console
435-/u-boot/lcd9b/.bm-work/00/common/lcd.c:120: undefined reference to `flush_dcache_range'
436+/u-boot/lcd9b/.bm-work/00/common/lcd.c:125: undefined reference to `flush_dcache_range'
43717: tegra: Enable display/lcd support on Seaboard
43818: wip
439
440So the problem is in lcd.c, due to missing cache operations. This information
441should be enough to work out what that commit is doing to break these
442boards. (In this case pxa did not have cache operations defined).
443
444If you see error lines marked with - that means that the errors were fixed
445by that commit. Sometimes commits can be in the wrong order, so that a
446breakage is introduced for a few commits and fixed by later commits. This
447shows up clearly with buildman. You can then reorder the commits and try
448again.
449
450At commit 16, the error moves - you can see that the old error at line 120
451is fixed, but there is a new one at line 126. This is probably only because
Dirk Behme3e1ded12014-12-23 07:41:26 +0100452we added some code and moved the broken line further down the file.
Simon Glassfc3fe1c2013-04-03 11:07:16 +0000453
454If many boards have the same error, then -e will display the error only
Simon Glassed966652014-08-28 09:43:43 -0600455once. This makes the output as concise as possible. To see which boards have
456each error, use -l.
Simon Glassfc3fe1c2013-04-03 11:07:16 +0000457
Simon Glasse30965d2014-08-28 09:43:44 -0600458Buildman tries to distinguish warnings from errors, and shows warning lines
459separately with a 'w' prefix.
460
Simon Glassfc3fe1c2013-04-03 11:07:16 +0000461The full build output in this case is available in:
462
463../lcd9b/12_of_18_gd92aff7_lcd--Add-support-for/lubbock/
464
465 done: Indicates the build was done, and holds the return code from make.
466 This is 0 for a good build, typically 2 for a failure.
467
468 err: Output from stderr, if any. Errors and warnings appear here.
469
470 log: Output from stdout. Normally there isn't any since buildman runs
471 in silent mode for now.
472
473 toolchain: Shows information about the toolchain used for the build.
474
475 sizes: Shows image size information.
476
477It is possible to get the build output there also. Use the -k option for
478this. In that case you will also see some output files, like:
479
480 System.map toolchain u-boot u-boot.bin u-boot.map autoconf.mk
481 (also SPL versions u-boot-spl and u-boot-spl.bin if available)
482
483
484Checking Image Sizes
485====================
486
487A key requirement for U-Boot is that you keep code/data size to a minimum.
488Where a new feature increases this noticeably it should normally be put
489behind a CONFIG flag so that boards can leave it off and keep the image
490size more or less the same with each new release.
491
492To check the impact of your commits on image size, use -S. For example:
493
494$ ./tools/buildman/buildman -b us-x86 -sS
495Summary of 10 commits for 1066 boards (4 threads, 1 job per thread)
49601: MAKEALL: add support for per architecture toolchains
49702: x86: Add function to get top of usable ram
498 x86: (for 1/3 boards) text -272.0 rodata +41.0
49903: x86: Add basic cache operations
50004: x86: Permit bootstage and timer data to be used prior to relocation
501 x86: (for 1/3 boards) data +16.0
50205: x86: Add an __end symbol to signal the end of the U-Boot binary
503 x86: (for 1/3 boards) text +76.0
50406: x86: Rearrange the output input to remove BSS
505 x86: (for 1/3 boards) bss -2140.0
50607: x86: Support relocation of FDT on start-up
507 x86: + coreboot-x86
50808: x86: Add error checking to x86 relocation code
50909: x86: Adjust link device tree include file
51010: x86: Enable CONFIG_OF_CONTROL on coreboot
511
512
513You can see that image size only changed on x86, which is good because this
514series is not supposed to change any other board. From commit 7 onwards the
515build fails so we don't get code size numbers. The numbers are fractional
516because they are an average of all boards for that architecture. The
517intention is to allow you to quickly find image size problems introduced by
518your commits.
519
520Note that the 'text' region and 'rodata' are split out. You should add the
521two together to get the total read-only size (reported as the first column
522in the output from binutil's 'size' utility).
523
524A useful option is --step which lets you skip some commits. For example
525--step 2 will show the image sizes for only every 2nd commit (so it will
526compare the image sizes of the 1st, 3rd, 5th... commits). You can also use
527--step 0 which will compare only the first and last commits. This is useful
528for an overview of how your entire series affects code size.
529
530You can also use -d to see a detailed size breakdown for each board. This
531list is sorted in order from largest growth to largest reduction.
532
533It is possible to go a little further with the -B option (--bloat). This
Simon Glasscec83c32014-08-09 15:32:57 -0600534shows where U-Boot has bloated, breaking the size change down to the function
Simon Glassfc3fe1c2013-04-03 11:07:16 +0000535level. Example output is below:
536
537$ ./tools/buildman/buildman -b us-mem4 -sSdB
538...
53919: Roll crc32 into hash infrastructure
540 arm: (for 10/10 boards) all -143.4 bss +1.2 data -4.8 rodata -48.2 text -91.6
541 paz00 : all +23 bss -4 rodata -29 text +56
542 u-boot: add: 1/0, grow: 3/-2 bytes: 168/-104 (64)
543 function old new delta
544 hash_command 80 160 +80
545 crc32_wd_buf - 56 +56
546 ext4fs_read_file 540 568 +28
547 insert_var_value_sub 688 692 +4
548 run_list_real 1996 1992 -4
549 do_mem_crc 168 68 -100
550 trimslice : all -9 bss +16 rodata -29 text +4
551 u-boot: add: 1/0, grow: 1/-3 bytes: 136/-124 (12)
552 function old new delta
553 hash_command 80 160 +80
554 crc32_wd_buf - 56 +56
555 ext4fs_iterate_dir 672 668 -4
556 ext4fs_read_file 568 548 -20
557 do_mem_crc 168 68 -100
558 whistler : all -9 bss +16 rodata -29 text +4
559 u-boot: add: 1/0, grow: 1/-3 bytes: 136/-124 (12)
560 function old new delta
561 hash_command 80 160 +80
562 crc32_wd_buf - 56 +56
563 ext4fs_iterate_dir 672 668 -4
564 ext4fs_read_file 568 548 -20
565 do_mem_crc 168 68 -100
566 seaboard : all -9 bss -28 rodata -29 text +48
567 u-boot: add: 1/0, grow: 3/-2 bytes: 160/-104 (56)
568 function old new delta
569 hash_command 80 160 +80
570 crc32_wd_buf - 56 +56
571 ext4fs_read_file 548 568 +20
572 run_list_real 1996 2000 +4
573 do_nandboot 760 756 -4
574 do_mem_crc 168 68 -100
575 colibri_t20_iris: all -9 rodata -29 text +20
576 u-boot: add: 1/0, grow: 2/-3 bytes: 140/-112 (28)
577 function old new delta
578 hash_command 80 160 +80
579 crc32_wd_buf - 56 +56
580 read_abs_bbt 204 208 +4
581 do_nandboot 760 756 -4
582 ext4fs_read_file 576 568 -8
583 do_mem_crc 168 68 -100
584 ventana : all -37 bss -12 rodata -29 text +4
585 u-boot: add: 1/0, grow: 1/-3 bytes: 136/-124 (12)
586 function old new delta
587 hash_command 80 160 +80
588 crc32_wd_buf - 56 +56
589 ext4fs_iterate_dir 672 668 -4
590 ext4fs_read_file 568 548 -20
591 do_mem_crc 168 68 -100
592 harmony : all -37 bss -16 rodata -29 text +8
593 u-boot: add: 1/0, grow: 2/-3 bytes: 140/-124 (16)
594 function old new delta
595 hash_command 80 160 +80
596 crc32_wd_buf - 56 +56
597 nand_write_oob_syndrome 428 432 +4
598 ext4fs_iterate_dir 672 668 -4
599 ext4fs_read_file 568 548 -20
600 do_mem_crc 168 68 -100
601 medcom-wide : all -417 bss +28 data -16 rodata -93 text -336
602 u-boot: add: 1/-1, grow: 1/-2 bytes: 88/-376 (-288)
603 function old new delta
604 crc32_wd_buf - 56 +56
605 do_fat_read_at 2872 2904 +32
606 hash_algo 16 - -16
607 do_mem_crc 168 68 -100
608 hash_command 420 160 -260
609 tec : all -449 bss -4 data -16 rodata -93 text -336
610 u-boot: add: 1/-1, grow: 1/-2 bytes: 88/-376 (-288)
611 function old new delta
612 crc32_wd_buf - 56 +56
613 do_fat_read_at 2872 2904 +32
614 hash_algo 16 - -16
615 do_mem_crc 168 68 -100
616 hash_command 420 160 -260
617 plutux : all -481 bss +16 data -16 rodata -93 text -388
618 u-boot: add: 1/-1, grow: 1/-3 bytes: 68/-408 (-340)
619 function old new delta
620 crc32_wd_buf - 56 +56
621 do_load_serial_bin 1688 1700 +12
622 hash_algo 16 - -16
623 do_fat_read_at 2904 2872 -32
624 do_mem_crc 168 68 -100
625 hash_command 420 160 -260
626 powerpc: (for 5/5 boards) all +37.4 data -3.2 rodata -41.8 text +82.4
627 MPC8610HPCD : all +55 rodata -29 text +84
628 u-boot: add: 1/0, grow: 0/-1 bytes: 176/-96 (80)
629 function old new delta
630 hash_command - 176 +176
631 do_mem_crc 184 88 -96
632 MPC8641HPCN : all +55 rodata -29 text +84
633 u-boot: add: 1/0, grow: 0/-1 bytes: 176/-96 (80)
634 function old new delta
635 hash_command - 176 +176
636 do_mem_crc 184 88 -96
637 MPC8641HPCN_36BIT: all +55 rodata -29 text +84
638 u-boot: add: 1/0, grow: 0/-1 bytes: 176/-96 (80)
639 function old new delta
640 hash_command - 176 +176
641 do_mem_crc 184 88 -96
642 sbc8641d : all +55 rodata -29 text +84
643 u-boot: add: 1/0, grow: 0/-1 bytes: 176/-96 (80)
644 function old new delta
645 hash_command - 176 +176
646 do_mem_crc 184 88 -96
647 xpedite517x : all -33 data -16 rodata -93 text +76
648 u-boot: add: 1/-1, grow: 0/-1 bytes: 176/-112 (64)
649 function old new delta
650 hash_command - 176 +176
651 hash_algo 16 - -16
652 do_mem_crc 184 88 -96
653...
654
655
656This shows that commit 19 has increased text size for arm (although only one
657board was built) and by 96 bytes for powerpc. This increase was offset in both
658cases by reductions in rodata and data/bss.
659
Dirk Behme3e1ded12014-12-23 07:41:26 +0100660Shown below the summary lines are the sizes for each board. Below each board
661are the sizes for each function. This information starts with:
Simon Glassfc3fe1c2013-04-03 11:07:16 +0000662
663 add - number of functions added / removed
664 grow - number of functions which grew / shrunk
665 bytes - number of bytes of code added to / removed from all functions,
666 plus the total byte change in brackets
667
668The change seems to be that hash_command() has increased by more than the
669do_mem_crc() function has decreased. The function sizes typically add up to
670roughly the text area size, but note that every read-only section except
671rodata is included in 'text', so the function total does not exactly
672correspond.
673
674It is common when refactoring code for the rodata to decrease as the text size
675increases, and vice versa.
676
677
Simon Glass62005342014-12-01 17:34:02 -0700678The .buildman file
679==================
Simon Glass4281ad82013-09-23 17:35:17 -0600680
Simon Glass62005342014-12-01 17:34:02 -0700681The .buildman file provides information about the available toolchains and
682also allows build flags to be passed to 'make'. It consists of several
683sections, with the section name in square brackets. Within each section are
684a set of (tag, value) pairs.
Simon Glass4281ad82013-09-23 17:35:17 -0600685
Simon Glass62005342014-12-01 17:34:02 -0700686'[toolchain]' section
Simon Glass4281ad82013-09-23 17:35:17 -0600687
Simon Glass62005342014-12-01 17:34:02 -0700688 This lists the available toolchains. The tag here doesn't matter, but
689 make sure it is unique. The value is the path to the toolchain. Buildman
690 will look in that path for a file ending in 'gcc'. It will then execute
691 it to check that it is a C compiler, passing only the --version flag to
692 it. If the return code is 0, buildman assumes that it is a valid C
693 compiler. It uses the first part of the name as the architecture and
694 strips off the last part when setting the CROSS_COMPILE environment
695 variable (parts are delimited with a hyphen).
Simon Glass4281ad82013-09-23 17:35:17 -0600696
Simon Glass62005342014-12-01 17:34:02 -0700697 For example powerpc-linux-gcc will be noted as a toolchain for 'powerpc'
698 and CROSS_COMPILE will be set to powerpc-linux- when using it.
699
700'[toolchain-alias]' section
701
702 This converts toolchain architecture names to U-Boot names. For example,
703 if an x86 toolchains is called i386-linux-gcc it will not normally be
704 used for architecture 'x86'. Adding 'x86: i386' to this section will
705 tell buildman that the i386 toolchain can be used for x86.
706
707'[make-flags]' section
708
709 U-Boot's build system supports a few flags (such as BUILD_TAG) which
710 affect the build product. These flags can be specified in the buildman
711 settings file. They can also be useful when building U-Boot against other
712 open source software.
713
714 [make-flags]
715 at91-boards=ENABLE_AT91_TEST=1
716 snapper9260=${at91-boards} BUILD_TAG=442
717 snapper9g45=${at91-boards} BUILD_TAG=443
718
719 This will use 'make ENABLE_AT91_TEST=1 BUILD_TAG=442' for snapper9260
720 and 'make ENABLE_AT91_TEST=1 BUILD_TAG=443' for snapper9g45. A special
721 variable ${target} is available to access the target name (snapper9260
722 and snapper9g20 in this case). Variables are resolved recursively. Note
723 that variables can only contain the characters A-Z, a-z, 0-9, hyphen (-)
724 and underscore (_).
725
726 It is expected that any variables added are dealt with in U-Boot's
727 config.mk file and documented in the README.
728
729 Note that you can pass ad-hoc options to the build using environment
730 variables, for example:
731
732 SOME_OPTION=1234 ./tools/buildman/buildman my_board
Simon Glass4281ad82013-09-23 17:35:17 -0600733
734
Simon Glasse5a0e5d2014-08-09 15:33:03 -0600735Quick Sanity Check
736==================
737
738If you have made changes and want to do a quick sanity check of the
Simon Glass1d8104f2014-10-16 01:05:56 -0600739currently checked-out source, run buildman without the -b flag. This will
740build the selected boards and display build status as it runs (i.e. -v is
741enabled automatically). Use -e to see errors/warnings as well.
Simon Glasse5a0e5d2014-08-09 15:33:03 -0600742
743
Simon Glass5abab202014-12-01 17:33:57 -0700744Building Ranges
745===============
746
747You can build a range of commits by specifying a range instead of a branch
748when using the -b flag. For example:
749
750 upstream/master..us-buildman
751
752will build commits in us-buildman that are not in upstream/master.
753
754
Simon Glassfc3fe1c2013-04-03 11:07:16 +0000755Other options
756=============
757
758Buildman has various other command line options. Try --help to see them.
759
Simon Glass2c3deb92014-08-28 09:43:39 -0600760When doing builds, Buildman's return code will reflect the overall result:
761
762 0 (success) No errors or warnings found
763 128 Errors found
764 129 Warnings found
765
Simon Glassfc3fe1c2013-04-03 11:07:16 +0000766
Simon Glass6eede342014-08-09 15:32:58 -0600767How to change from MAKEALL
768==========================
769
770Buildman includes most of the features of MAKEALL and is generally faster
771and easier to use. In particular it builds entire branches: if a particular
772commit introduces an error in a particular board, buildman can easily show
773you this, even if a later commit fixes that error.
774
775The reasons to deprecate MAKEALL are:
776- We don't want to maintain two build systems
777- Buildman is typically faster
778- Buildman has a lot more features
779
780But still, many people will be sad to lose MAKEALL. If you are used to
781MAKEALL, here are a few pointers.
782
783First you need to set up your tool chains - see the 'Setting up' section
784for details. Once you have your required toolchain(s) detected then you are
785ready to go.
786
Simon Glasse5a0e5d2014-08-09 15:33:03 -0600787To build the current source tree, run buildman without a -b flag:
788
789 ./tools/buildman/buildman <list of things to build>
790
791This will build the current source tree for the given boards and display
792the results and errors.
793
794However buildman usually works on entire branches, and for that you must
795specify a board flag:
Simon Glass6eede342014-08-09 15:32:58 -0600796
797 ./tools/buildman/buildman -b <branch_name> <list of things to build>
798
799followed by (afterwards, or perhaps concurrently in another terminal):
800
801 ./tools/buildman/buildman -b <branch_name> -s <list of things to build>
802
803to see the results of the build. Rather than showing you all the output,
804buildman just shows a summary, with red indicating that a commit introduced
805an error and green indicating that a commit fixed an error. Use the -e
Simon Glassed966652014-08-28 09:43:43 -0600806flag to see the full errors and -l to see which boards caused which errors.
Simon Glass6eede342014-08-09 15:32:58 -0600807
Simon Glasse5a0e5d2014-08-09 15:33:03 -0600808If you really want to see build results as they happen, use -v when doing a
Simon Glass1d8104f2014-10-16 01:05:56 -0600809build (and -e to see the errors/warnings too).
Simon Glasse5a0e5d2014-08-09 15:33:03 -0600810
Simon Glass6eede342014-08-09 15:32:58 -0600811You don't need to stick around on that branch while buildman is running. It
812checks out its own copy of the source code, so you can change branches,
813add commits, etc. without affecting the build in progress.
814
815The <list of things to build> can include board names, architectures or the
816like. There are no flags to disambiguate since ambiguities are rare. Using
817the examples from MAKEALL:
818
819Examples:
820 - build all Power Architecture boards:
821 MAKEALL -a powerpc
822 MAKEALL --arch powerpc
823 MAKEALL powerpc
824 ** buildman -b <branch> powerpc
825 - build all PowerPC boards manufactured by vendor "esd":
826 MAKEALL -a powerpc -v esd
827 ** buildman -b <branch> esd
828 - build all PowerPC boards manufactured either by "keymile" or "siemens":
829 MAKEALL -a powerpc -v keymile -v siemens
830 ** buildman -b <branch> keymile siemens
831 - build all Freescale boards with MPC83xx CPUs, plus all 4xx boards:
832 MAKEALL -c mpc83xx -v freescale 4xx
833 ** buildman -b <branch> mpc83xx freescale 4xx
834
835Buildman automatically tries to use all the CPUs in your machine. If you
836are building a lot of boards it will use one thread for every CPU core
837it detects in your machine. This is like MAKEALL's BUILD_NBUILDS option.
838You can use the -T flag to change the number of threads. If you are only
839building a few boards, buildman will automatically run make with the -j
840flag to increase the number of concurrent make tasks. It isn't normally
841that helpful to fiddle with this option, but if you use the BUILD_NCPUS
842option in MAKEALL then -j is the equivalent in buildman.
843
844Buildman puts its output in ../<branch_name> by default but you can change
845this with the -o option. Buildman normally does out-of-tree builds: use -i
846to disable that if you really want to. But be careful that once you have
847used -i you pollute buildman's copies of the source tree, and you will need
848to remove the build directory (normally ../<branch_name>) to run buildman
849in normal mode (without -i).
850
851Buildman doesn't keep the output result normally, but use the -k option to
852do this.
853
854Please read 'Theory of Operation' a few times as it will make a lot of
855things clearer.
856
857Some options you might like are:
858
859 -B shows which functions are growing/shrinking in which commit - great
860 for finding code bloat.
861 -S shows image sizes for each commit (just an overall summary)
862 -u shows boards that you haven't built yet
863 --step 0 will build just the upstream commit and the last commit of your
864 branch. This is often a quick sanity check that your branch doesn't
865 break anything. But note this does not check bisectability!
866
867
Simon Glassfc3fe1c2013-04-03 11:07:16 +0000868TODO
869====
870
871This has mostly be written in my spare time as a response to my difficulties
872in testing large series of patches. Apart from tidying up there is quite a
Simon Glass1d8104f2014-10-16 01:05:56 -0600873bit of scope for improvement. Things like better error diffs and easier
Dirk Behme3e1ded12014-12-23 07:41:26 +0100874access to log files. Also it would be nice if buildman could 'hunt' for
Simon Glass1d8104f2014-10-16 01:05:56 -0600875problems, perhaps by building a few boards for each arch, or checking
876commits for changed files and building only boards which use those files.
Simon Glassfc3fe1c2013-04-03 11:07:16 +0000877
878
879Credits
880=======
881
882Thanks to Grant Grundler <grundler@chromium.org> for his ideas for improving
883the build speed by building all commits for a board instead of the other
884way around.
885
886
Simon Glassfc3fe1c2013-04-03 11:07:16 +0000887Simon Glass
888sjg@chromium.org
889Halloween 2012
890Updated 12-12-12
891Updated 23-02-13