git filter-branch花了很长时间。令人高兴的是，我找到了BFG repo-cleaner。
但它意外地改变了我上一次提交的内容。
$ git clone --mirror example.com:/repo.git
$ cd repo.git
$ git log HEAD^!
commit 5f737d28756d4854d25899632abffe7cca2c7423
Author: Paul Draper <paul@example.com>
Date: Sat Jan 24 19:31:47 2015 -0700
Fix /contact and /folderEntries/listFoldersSimple
$ git diff --stat HEAD^!
cake/app/controllers/folder_entries_controller.php | 1 +
现在我干净了。
$ java -jar ~/bfg-1.12.0.jar -b 1M
...
In total, 161797 object ids were changed. Full details are logged here:
...
$ git log HEAD^!
commit 3ff700cebe32497423435b416ea11169b7fcbf90
Author: Paul Draper <paul@example.com>
Date: Sat Jan 24 19:31:47 2015 -0700
Fix /contact and /folderEntries/listFoldersSimple
Former-commit-id: 5f737d28756d4854d25899632abffe7cca2c7423
$ git diff --stat HEAD^!
cake/app/controllers/folder_entries_controller.php | 1 +
.../lucidchart-tools/caja/ant-jars/guava-r09.jar | Bin 0 -> 1141964 bytes
.../caja/ant-jars/guava-r09.jar.REMOVED.git-id | 1 -
cake/app/lucidchart-tools/caja/ant-jars/js.jar | Bin 0 -> 1122370 bytes
.../caja/ant-jars/js.jar.REMOVED.git-id | 1 -
.../lucidchart-tools/caja/ant-jars/pluginc-src.jar | Bin 0 -> 5172676 bytes
.../caja/ant-jars/pluginc-src.jar.REMOVED.git-id | 1 -
.../app/lucidchart-tools/caja/ant-jars/pluginc.jar | Bin 0 -> 2959487 bytes
.../caja/ant-jars/pluginc.jar.REMOVED.git-id | 1 -
.../lucidchart-tools/caja/ant-jars/xercesImpl.jar | Bin 0 -> 1229125 bytes
.../caja/ant-jars/xercesImpl.jar.REMOVED.git-id | 1 -
cake/app/lucidchart-tools/jsdoc/rhino/js.jar | Bin 0 -> 1111429 bytes
.../jsdoc/rhino/js.jar.REMOVED.git-id | 1 -
cake/app/lucidchart-tools/selenium/chromedriver | Bin 0 -> 5778064 bytes
.../selenium/chromedriver.REMOVED.git-id | 1 -
.../selenium/selenium-server-standalone-2.37.0.jar | Bin 0 -> 34730734 bytes
...ium-server-standalone-2.37.0.jar.REMOVED.git-id | 1 -
.../selenium-server-standalone-2.42.2-mod.jar | Bin 0 -> 34873583 bytes
...server-standalone-2.42.2-mod.jar.REMOVED.git-id | 1 -
.../selenium/selenium-server-standalone-2.42.2.jar | Bin 0 -> 34823352 bytes
...ium-server-standalone-2.42.2.jar.REMOVED.git-id | 1 -
.../lucidchart-tools/test-runner-1.0-SNAPSHOT.jar | Bin 0 -> 9732125 bytes
.../test-runner-1.0-SNAPSHOT.jar.REMOVED.git-id | 1 -
.../CommandLine/Scaffolders/DefaultScaffolder.phar | Bin 0 -> 4404199 bytes
.../DefaultScaffolder.phar.REMOVED.git-id | 1 -
.../WebPICmdLine/Microsoft.Web.Deployment.dll | Bin 0 -> 1201991 bytes
.../Microsoft.Web.Deployment.dll.REMOVED.git-id | 1 -
cake/app/vendors/aws.phar | Bin 0 -> 6784935 bytes
cake/app/vendors/aws.phar.REMOVED.git-id | 1 -
.../tcpdf/fonts/dejavu-fonts-ttf-2.33/status.txt | 6657 +++++
.../status.txt.REMOVED.git-id | 1 -
cake/app/vendors/tcpdf/tcpdf.php | 28808 +++++++++++++++++++
cake/app/vendors/tcpdf/tcpdf.php.REMOVED.git-id | 1 -
.../img/onboarding-chart/04_shape manager.gif | Bin 0 -> 1413721 bytes
.../04_shape manager.gif.REMOVED.git-id | 1 -
cake/app/webroot/img/onboarding-chart/05_share.gif | Bin 0 -> 1341876 bytes
.../onboarding-chart/05_share.gif.REMOVED.git-id | 1 -
.../js/closure/usage/rhino/javadoc/index-all.html | 12027 ++++++++
.../rhino/javadoc/index-all.html.REMOVED.git-id | 1 -
cake/app/webroot/js/closure/usage/rhino/js-14.jar | Bin 0 -> 1471932 bytes
.../closure/usage/rhino/js-14.jar.REMOVED.git-id | 1 -
cake/app/webroot/js/closure/usage/rhino/js.jar | Bin 0 -> 1134765 bytes
.../js/closure/usage/rhino/js.jar.REMOVED.git-id | 1 -
.../js/closure/usage/rhino/testsrc/tests.tar.gz | Bin 0 -> 1778543 bytes
.../rhino/testsrc/tests.tar.gz.REMOVED.git-id | 1 -
cake/app/webroot/js/mathquill/font/Symbola.svg | 5102 ++++
.../js/mathquill/font/Symbola.svg.REMOVED.git-id | 1 -
.../webroot/js/templates/SoyToJsSrcCompiler.jar | Bin 0 -> 2154164 bytes
.../SoyToJsSrcCompiler.jar.REMOVED.git-id | 1 -
cake/app/webroot/persona-pages/img/gif-v3.gif | Bin 0 -> 1570363 bytes
.../persona-pages/img/gif-v3.gif.REMOVED.git-id | 1 -
.../webroot/persona-pages/img/interactive-gif.gif | Bin 0 -> 1434134 bytes
.../img/interactive-gif.gif.REMOVED.git-id | 1 -
cake/build/closure/compiler.jar | Bin 0 -> 6007184 bytes
cake/build/closure/compiler.jar.REMOVED.git-id | 1 -
.../lucidchart-mobile-sliders-landscape-4.png | Bin 0 -> 1718536 bytes
...t-mobile-sliders-landscape-4.png.REMOVED.git-id | 1 -
.../lucidchart-mobile-sliders-portrait-4.png | Bin 0 -> 1614308 bytes
...rt-mobile-sliders-portrait-4.png.REMOVED.git-id | 1 -
.../Versions/A/OCHamcrestIOS | Bin 0 -> 3671740 bytes
.../Versions/A/OCHamcrestIOS.REMOVED.git-id | 1 -
.../OCMockitoIOS.framework/Versions/A/OCMockitoIOS | Bin 0 -> 1299132 bytes
.../Versions/A/OCMockitoIOS.REMOVED.git-id | 1 -
.../Versions/A/CrashReporter | Bin 0 -> 1432156 bytes
.../Versions/A/CrashReporter.REMOVED.git-id | 1 -
chart-ios/libFlurry_6.0.0.a | Bin 0 -> 3819300 bytes
chart-ios/libFlurry_6.0.0.a.REMOVED.git-id | 1 -
67 files changed, 52595 insertions(+), 33 deletions(-)
所有这些额外文件都是我想删除的文件。
为什么在我最近的提交中更改了所有这些文件？
答案 0 :(得分：6)
显然，这是关于BFG如何运作的常见误解。来自documentation：
如果有问题 - 比如一个10MB的文件，当你告诉BFG去掉每个超过5MB的文件时 - 在受保护的提交中，它将不会被删除，而因为它仍然在那里，所以没有从早先的提交中删除它。如果你想要BFG删除你需要确保当前提交是干净的东西。
这并不意味着“没有必要先删除它，所以它不会”，这就是我所读到的。
这意味着“没有必要先删除它，所以它可能不会”。在任何情况下，它都将遵守受保护的提交具有相同树的保证。
在我的情况下，它确实先删除了这些blob，但后来不得不重新添加内容以保留HEAD树不变的要求。
在Github上有关于此的更完整的讨论：
修改
我找到了一种方法来移除HEAD上存在的除之外的大blob。
这使用bash + unix实用程序来查找超过1MB的任何blob（对于不同的大小更改1024 * 1024），然后使用BFG将其删除：
comm -23 \
<(git rev-list --objects --all | git cat-file --batch-check="%(objecttype) %(objectname) %(objectsize) %(rest)" | grep ^blob | awk '$3 > 1024 * 1024 { print $2 }' | sort) \
<(git ls-tree -r HEAD | cut -f 1 | cut -d ' ' -f 3 | sort) \
> /tmp/large-blobs.list
java -jar bfg-1.12.0.jar -bi /tmp/large-blobs.list