[AGP 3.6.0-beta01] SecurityException: SHA1 digest error for BouncyCastleProvider when using Robolectric [142580430]

Fixed

Bug

Status Update

No update yet.

Description

ch...@gmail.com

created issue #1

Oct 12, 2019 11:43PM

All of my Robolectric tests are now failing due to a SecurityException. These previously worked fine with AGP 3.6.0-alpha12.

Log:
java.lang.SecurityException: SHA1 digest error for org/bouncycastle/jce/provider/BouncyCastleProvider.class
at sun.security.util.ManifestEntryVerifier.verify(ManifestEntryVerifier.java:223)
at java.util.jar.JarVerifier.processEntry(JarVerifier.java:243)
at java.util.jar.JarVerifier.update(JarVerifier.java:230)
at java.util.jar.JarVerifier$VerifierStream.read(JarVerifier.java:484)
at java.io.FilterInputStream.read(FilterInputStream.java:133)
at java.io.FilterInputStream.read(FilterInputStream.java:107)
at org.robolectric.util.Util.copy(Util.java:46)
at org.robolectric.util.Util.readBytes(Util.java:63)
at org.robolectric.internal.bytecode.SandboxClassLoader.getByteCode(SandboxClassLoader.java:167)
at org.robolectric.internal.bytecode.SandboxClassLoader.maybeInstrumentClass(SandboxClassLoader.java:119)
at org.robolectric.internal.bytecode.SandboxClassLoader.lambda$findClass$0(SandboxClassLoader.java:112)
at org.robolectric.util.PerfStatsCollector.measure(PerfStatsCollector.java:53)
at org.robolectric.internal.bytecode.SandboxClassLoader.findClass(SandboxClassLoader.java:111)
at java.lang.ClassLoader.loadClass(ClassLoader.java:424)
at java.lang.ClassLoader.loadClass(ClassLoader.java:357)
at java.lang.Class.getDeclaredConstructors0(Native Method)
at java.lang.Class.privateGetDeclaredConstructors(Class.java:2671)
at java.lang.Class.getConstructors(Class.java:1651)
at org.robolectric.util.inject.Injector.findConstructor(Injector.java:249)
at org.robolectric.util.inject.Injector.inject(Injector.java:232)
at org.robolectric.util.inject.Injector.lambda$new$0(Injector.java:115)
at org.robolectric.util.inject.Injector$MemoizingProvider.get(Injector.java:485)
at org.robolectric.util.inject.Injector.getInstanceInternal(Injector.java:213)
at org.robolectric.util.inject.Injector.getInstance(Injector.java:197)
at org.robolectric.util.inject.Injector.getInstance(Injector.java:191)
at org.robolectric.internal.AndroidSandbox.lambda$new$0(AndroidSandbox.java:51)
at java.util.concurrent.FutureTask.run(FutureTask.java:266)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
at java.lang.Thread.run(Thread.java:748)

Repro:
git clone -b cb/agb-3.6.0-beta01

https://github.com/chrisbanes/tivi.git
cd tivi
./gradlew testDebug

Comments

ga...@google.com <ga...@google.com> Oct 14, 2019 10:51AM

Assigned to ga...@google.com.

ga...@google.com <ga...@google.com> #2Oct 14, 2019 02:39PM

what is in update.zip ? can you share that file to triage this issue

Message last modified on Oct 14, 2019 02:52PM

pa...@google.com <pa...@google.com> Oct 14, 2019 02:54PM

Reassigned to pa...@google.com.

ga...@google.com <ga...@google.com> #3Oct 14, 2019 03:12PM

MHC19Q for Nexus 5X (bullhead) -

https://dl.google.com/dl/android/aosp//bullhead-ota-mhc19q-8fe67a2b.zip
N4F26T for Nexus 5X (bullhead) -

https://dl.google.com/dl/android/aosp/bullhead-ota-n4f26t-648ce802.zip
WW-12.2.5.23 for Asus ZenFone Go (ASUS_X014D) -

http://dlcdnet.asus.com/pub/ASUS/ZenFone/ZB452KG/UL-ASUS_X014D-WW-12.2.5.23-user.zip

Look like, this happens only with a big images.

ch...@gmail.com <ch...@gmail.com> #4Oct 14, 2019 08:28PM

On Windows 10 (x64)

pa...@google.com <pa...@google.com> #5Oct 25, 2019 05:37PM

Accepted by pa...@google.com.

related to (and probably worked around for now by) internal bug

http://b/36046324

i do think we should also start shipping a 64-bit Windows platform tools package.

we should also consider changing android::base::ReadFdToString to call fstat and pre-size the vector. not worthwhile for reading things like /proc/uptime, but the default expansion behavior for a huge file like a full OTA update is going to result in substantial overhead. (which will explain the regression between .3 and .4, since the former will have allocated exactly the necessary number of bytes.)

lastly (and maybe not worth doing at all, depending on how soon we can get folks to use a 64-bit platform tools on Windows, since all the other platforms are already 64-bit-only), we could consider rewriting the adb sideload code to (a) mmap/munmap rather than actually read in to physical memory or (b) pread. i'm not sure what the performance impact of pread would be (especially on Windows where there is no pread equivalent).

mi...@aexp.com <mi...@aexp.com> #6Nov 25, 2019 11:18PM

Correction to #5: Windows does support pread() - it's in ReadFile()'s OVERLAPPED parameter.

pa...@google.com <pa...@google.com> #7Nov 26, 2019 02:33PM

@7: we'd still need to write the pread wrapper, though. (similar if we went with mmap/munmap.)

https://android-review.googlesource.com/#/c/355481/ pre-sizes the vector, which should be enough to undo the regression from '.3 to '.4...

hu...@google.com <hu...@google.com> #8Nov 27, 2019 05:29PM

Marked as fixed.

https://android-review.googlesource.com/355818 switches "adb sideload" over to lseek/read. the time difference isn't measurable, but the memory requirements are then down to just 64KiB.