This is an archive of the discontinued LLVM Phabricator instance.

Differential D66015

[llvm-strings] Improve testing of llvm-strings
ClosedPublic

Authored by jhenderson on Aug 9 2019, 8:20 AM.

Details

Reviewers
rupprecht
grimar
MaskRay
compnerd
Commits
rGf23ce128fd61: [llvm-strings] Improve testing of llvm-strings
rL368555: [llvm-strings] Improve testing of llvm-strings
Summary

This patch attempts to tidy up the llvm-strings testing by:

  1. Adding comments to every test.
  2. Getting rid of canned input files, and have the tests generate them on the fly (this makes the tests self-contained).
  3. Adding missing test coverage.
  4. Renaming some tests that weren't clear as to their purpose.
  5. Adding extra checking of various cases, formatting etc.
  6. Removing a test that didn't seem to have any useful purpose for testing llvm-strings.

Diff Detail

Repository
rL LLVM

Event Timeline

jhenderson created this revision.Aug 9 2019, 8:20 AM
Herald added a project: Restricted Project. · View Herald TranscriptAug 9 2019, 8:20 AM
rupprecht accepted this revision.Aug 9 2019, 8:54 AM

Just one test that needs fixing, LGTM otherwise:

test/tools/llvm-strings/version.test
4

This is actually testing the default cmake config; see this comment from llvm/test/tools/llvm-readobj/basic.test:

# Test version switch.
RUN: llvm-readobj --version | FileCheck %s --check-prefix=VERSION
RUN: llvm-readelf --version | FileCheck %s --check-prefix=VERSION
# In default configuration we could match "LLVM version", but the "LLVM" part
# can be changed with PACKAGE_NAME in CMake, so we match only version.
VERSION: version
This revision is now accepted and ready to land.Aug 9 2019, 8:54 AM
MaskRay accepted this revision.Aug 9 2019, 9:56 PM
grimar added inline comments.Aug 10 2019, 5:42 AM
test/tools/llvm-strings/all-sections.test
1

Should we use ## in this tests too? I would do this, since it is probably
not bad to have a single style across the all LLVM tests.

test/tools/llvm-strings/eof-no-string.test
2

I think it just does not print all strings that are shorter than 4 bytes?
I.e. not only the last line.

test/tools/llvm-strings/eof.test
5

What about combining this and eof-no-string.test test into one?

test/tools/llvm-strings/file-filename.test
10

My emotion for this test at first was: "what about multiple inputs?". Then I found
multiple-inputs.test test below. I would just inline it here.

test/tools/llvm-strings/help.test
12

Note: I am not sure why "Options" is in upper case when "Generic", but lower case when "General",
but my concern is that it makes the test case to be potentially brittle probably:

If you have --implicit-check-not="Generic Options:", then if somebody deside to change "Option"->"option",
then you'll get a broken test case. Not sure if we should think about this scenario in that way though.

test/tools/llvm-strings/length.test
13

What if -n 6 or -n -1 or -n foo?

test/tools/llvm-strings/version.test
1

I.d probably just put this into help.test. What do you think?

jhenderson marked 10 inline comments as done.Aug 12 2019, 3:07 AM
jhenderson added inline comments.
test/tools/llvm-strings/all-sections.test
1

I didn't bother here because this isn't used as an input to anything, but I can do so if you're bothered. Should I add '#' to the RUN and CHECK lines too? I'd prefer not to, mostly because that will unnecessarily make git blame less useful.

test/tools/llvm-strings/eof-no-string.test
1

I'm not sure why my patch isn't showing this, but this has been renamed from "terminator-neg.test".

2

Right, but this is specifically about the last line case (there are other cases for the general behaviour). A basic implementation would look something like:

const char *S = nullptr;
for (const char *P = Str.begin(); P != Str.end(); ++P) {
  if (!isPrint(P)) {
    if (P - S >= 4)
      printStr(S, P);
    S = nullptr;
  }
}

This wouldn't print the trailing characters at all. A naive fix would add the following after the loop:

if (S != nullptr)
  printStr(S, Str.end());

When the check should actually be if (S != nullptr && Str.end() - S >= 4). So this test case is checking that second clause in the final if.

Note that this and the original eof.test were written to fix a bug where that final if was missing.

test/tools/llvm-strings/eof.test
1

I'm not sure why my patch isn't showing this, but this has been renamed from "terminator.test".

5

I could do that, but I kept them separate because that's how they were previously. If you'd like me to fold them together, I can do so.

test/tools/llvm-strings/file-filename.test
10

Okay.

test/tools/llvm-strings/help.test
12

I agree that the inconsistent case is bad. I'm not sure of a good alternative (apart from making it a regex pattern of [Oo]). I'll give it a bit more thought.

test/tools/llvm-strings/length.test
13

-n -1 is actually broken. It looks like the command-line parser converts it to an unsigned, resulting in it being the equivalent of UINT_MAX. As GNU strings allows -1 as an alias for -n 1, we probably don't want to try to address this immediately, and fix that issue first, then see if the problem is still possible afterwards.

I'll add -n foo and -n 6 cases.

test/tools/llvm-strings/version.test
1

I prefer to keep each switch to a separate test, unless they're intricately linked (in this case --help and --version are not really linked).

4

Thanks! I'll fix that.

jhenderson marked an inline comment as done.Aug 12 2019, 3:18 AM
jhenderson added inline comments.
test/tools/llvm-strings/file-filename.test
10

Actually, multiple-inputs.test isn't just about the --print-file-name switch. It's more generally about how llvm-strings works with multiple inputs, i.e. it reads each of them in turn, and produces correct output with --print-filename and --radix when switching between files, so I think it belongs as a separate test.

Possibly I could be persuaded that the --print-file-name and --radix checks for multiple inputs belong in their respective tests, but I think multiple inputs still need testing separately.

grimar added inline comments.Aug 12 2019, 3:22 AM
test/tools/llvm-strings/all-sections.test
1

I'd use '##' fo comments, I agree it is probably no need to use them for RUN/CHECK lines.

test/tools/llvm-strings/eof.test
5

Phab did not show the renaming, so I assumed this is the new file.
Since it is not, it is OK to leave it as is in this patch I think.

(I would fold them in a followup though).

jhenderson updated this revision to Diff 214596.Aug 12 2019, 3:39 AM
jhenderson marked 8 inline comments as done.

Addressed review comments:

  • Made comments use double hash.
  • Made some tests less brittle.
  • Added extra test cases as suggested.
test/tools/llvm-strings/help.test
12

I've tried making it less likely to get missed, by putting the -NOT checks out-of-line, so that they're next to the positive checks. If the case is changed, hopefully the person updating the test will change both positive and negative cases.

test/tools/llvm-strings/length.test
13

-n -1 is actually broken. It looks like the command-line parser converts it to an unsigned

Turns out it's an int in the parser, so the parser accepts it, but it is then static_cast to a size_t! Anyway, still broken, and I've filed a bug for negative arguments.

grimar accepted this revision.Aug 12 2019, 3:46 AM

LGTM.

jhenderson marked an inline comment as done.Aug 12 2019, 4:31 AM
jhenderson added inline comments.
test/tools/llvm-strings/stdin.test
12

On my Linux box, "echo -e" doesn't work without quoting the string to be echoed, so I did that throughout this test.

Closed by commit rL368555: [llvm-strings] Improve testing of llvm-strings (authored by jhenderson). · Explain WhyAug 12 2019, 4:37 AM
This revision was automatically updated to reflect the committed changes.
jhenderson mentioned this in D67069: [llvm-strings][test] Merge two closely-related tests.Sep 2 2019, 3:24 AM
jhenderson mentioned this in rL370643: [llvm-strings][test] Merge two closely related tests.Sep 2 2019, 4:44 AM
jhenderson mentioned this in rG43e9ead1edde: [llvm-strings][test] Merge two closely related tests.

Revision Contents

Diff 214381

test/tools/llvm-strings/Inputs/abcd

abcd
No newline at end of file

test/tools/llvm-strings/Inputs/numbers

one
two
three
four
five
six
seven
eight
nine
ten

test/tools/llvm-strings/Inputs/variable-length

a
ab
abc
abcd

test/tools/llvm-strings/all-sections.test

RUN: llvm-strings -a %S/Inputs/abcd | FileCheck %s # Show that the -a/--all switch is accepted. GNU strings checks specific object
grimarUnsubmitted
Done
ReplyInline Actions

Should we use ## in this tests too? I would do this, since it is probably
not bad to have a single style across the all LLVM tests.

grimar: Should we use `##` in this tests too? I would do this, since it is probably not bad to have a…
jhendersonAuthorUnsubmitted
Done
ReplyInline Actions

I didn't bother here because this isn't used as an input to anything, but I can do so if you're bothered. Should I add '#' to the RUN and CHECK lines too? I'd prefer not to, mostly because that will unnecessarily make git blame less useful.

jhenderson: I didn't bother here because this isn't used as an input to anything, but I can do so if you're…
grimarUnsubmitted
Done
ReplyInline Actions

I'd use '##' fo comments, I agree it is probably no need to use them for RUN/CHECK lines.

grimar: I'd use '##' fo comments, I agree it is probably no need to use them for `RUN`/`CHECK` lines.
RUN: llvm-strings --all %S/Inputs/abcd | FileCheck %s # file sections unless --all is specified. llvm-strings just looks at the entire
# file, so this option is silently ignored.
RUN: echo abcd | llvm-strings -a | FileCheck %s
RUN: echo abcd | llvm-strings --all | FileCheck %s
CHECK: abcd CHECK: abcd

test/tools/llvm-strings/archive-filename.test

# Show that --print-file-name prints just the archive name for an archive and
# not the member names.
RUN: rm -rf %t && mkdir -p %t RUN: rm -rf %t && mkdir -p %t
RUN: llvm-ar --format=gnu crs %t/archive.a %S/Inputs/abcd RUN: echo -n abcd > %t/abcd
RUN: llvm-ar --format=gnu crs %t/archive.a %t/abcd
RUN: llvm-strings -f %t/archive.a | FileCheck %s RUN: llvm-strings -f %t/archive.a | FileCheck %s
RUN: llvm-strings --print-file-name %t/archive.a | FileCheck %s RUN: llvm-strings --print-file-name %t/archive.a | FileCheck %s
CHECK: archive.a: !<arch> CHECK: archive.a: !<arch>
CHECK: archive.a: abcd/ 0 0 0 644 4 ` CHECK: archive.a: abcd/ 0 0 0 644 4 `
CHECK: archive.a: abcd CHECK: archive.a: abcd

test/tools/llvm-strings/eof-no-string.test

# Show that llvm-strings does not print the last string in the input if it is
jhendersonAuthorUnsubmitted
Done
ReplyInline Actions

I'm not sure why my patch isn't showing this, but this has been renamed from "terminator-neg.test".

jhenderson: I'm not sure why my patch isn't showing this, but this has been renamed from "terminator-neg.
# too short and no unprintable character follows it.
grimarUnsubmitted
Not Done
ReplyInline Actions

I think it just does not print all strings that are shorter than 4 bytes?
I.e. not only the last line.

grimar: I think it just does not print all strings that are shorter than 4 bytes? I.e. not only the…
jhendersonAuthorUnsubmitted
Done
ReplyInline Actions

Right, but this is specifically about the last line case (there are other cases for the general behaviour). A basic implementation would look something like:

const char *S = nullptr;
for (const char *P = Str.begin(); P != Str.end(); ++P) {
  if (!isPrint(P)) {
    if (P - S >= 4)
      printStr(S, P);
    S = nullptr;
  }
}

This wouldn't print the trailing characters at all. A naive fix would add the following after the loop:

if (S != nullptr)
  printStr(S, Str.end());

When the check should actually be if (S != nullptr && Str.end() - S >= 4). So this test case is checking that second clause in the final if.

Note that this and the original eof.test were written to fix a bug where that final if was missing.

jhenderson: Right, but this is specifically about the last line case (there are other cases for the general…
RUN: echo -n abc | llvm-strings - | FileCheck -allow-empty %s RUN: echo -n abc | llvm-strings - | FileCheck -allow-empty %s
CHECK-NOT: abc CHECK-NOT: abc

test/tools/llvm-strings/eof.test

# Show that llvm-strings prints the last string in the input even if no
jhendersonAuthorUnsubmitted
Done
ReplyInline Actions

I'm not sure why my patch isn't showing this, but this has been renamed from "terminator.test".

jhenderson: I'm not sure why my patch isn't showing this, but this has been renamed from "terminator.test".
# unprintable character follows it.
RUN: echo -n abcdefg | llvm-strings - | FileCheck %s RUN: echo -n abcdefg | llvm-strings - | FileCheck %s
CHECK: abcdefg CHECK: abcdefg
grimarUnsubmitted
Not Done
ReplyInline Actions

What about combining this and eof-no-string.test test into one?

grimar: What about combining this and `eof-no-string.test` test into one?
jhendersonAuthorUnsubmitted
Done
ReplyInline Actions

I could do that, but I kept them separate because that's how they were previously. If you'd like me to fold them together, I can do so.

jhenderson: I could do that, but I kept them separate because that's how they were previously. If you'd…
grimarUnsubmitted
Done
ReplyInline Actions

Phab did not show the renaming, so I assumed this is the new file.
Since it is not, it is OK to leave it as is in this patch I think.

(I would fold them in a followup though).

grimar: Phab did not show the renaming, so I assumed this is the new file. Since it is not, it is OK…

test/tools/llvm-strings/file-filename.test

RUN: llvm-strings -f %S/Inputs/abcd | FileCheck %s # Show that --print-file-name prints the name of the file before each line in
RUN: llvm-strings --print-file-name %S/Inputs/abcd | FileCheck %s --strict-whitespace # the output.
CHECK: {{[\\/]}}abcd: abcd
RUN: echo abcd > %t
RUN: echo efg >> %t
RUN: echo hijk >> %t
RUN: llvm-strings -f %t | FileCheck %s --strict-whitespace -DFILE=%t
RUN: llvm-strings --print-file-name %t | FileCheck %s --strict-whitespace -DFILE=%t
CHECK: [[FILE]]: abcd
CHECK-NEXT: [[FILE]]: hijk
grimarUnsubmitted
Not Done
ReplyInline Actions

My emotion for this test at first was: "what about multiple inputs?". Then I found
multiple-inputs.test test below. I would just inline it here.

grimar: My emotion for this test at first was: "what about multiple inputs?". Then I found `multiple…
jhendersonAuthorUnsubmitted
Done
ReplyInline Actions

Okay.

jhenderson: Okay.
jhendersonAuthorUnsubmitted
Done
ReplyInline Actions

Actually, multiple-inputs.test isn't just about the --print-file-name switch. It's more generally about how llvm-strings works with multiple inputs, i.e. it reads each of them in turn, and produces correct output with --print-filename and --radix when switching between files, so I think it belongs as a separate test.

Possibly I could be persuaded that the --print-file-name and --radix checks for multiple inputs belong in their respective tests, but I think multiple inputs still need testing separately.

jhenderson: Actually, multiple-inputs.test isn't just about the --print-file-name switch. It's more…

test/tools/llvm-strings/help.test

RUN: llvm-strings -h | FileCheck %s # Show that help text is printed correctly when requested.
RUN: llvm-strings --help | FileCheck %s
RUN: llvm-strings -h | FileCheck %s --check-prefixes=CHECK,CATEG
RUN: llvm-strings --help | FileCheck %s --check-prefixes=CHECK,CATEG
RUN: llvm-strings --help-list \
RUN: | FileCheck %s --implicit-check-not="General options:" --implicit-check-not="Generic Options:"
CHECK: OVERVIEW: llvm string dumper CHECK: OVERVIEW: llvm string dumper
CHECK: USAGE: llvm-strings{{(.exe)?}} [options] <input object files>{{$}} CHECK: USAGE: llvm-strings{{(.exe)?}} [options] <input object files>{{$}}
CHECK: OPTIONS: CHECK: OPTIONS:
CATEG: General options:
CATEG: Generic Options:
grimarUnsubmitted
Done
ReplyInline Actions

Note: I am not sure why "Options" is in upper case when "Generic", but lower case when "General",
but my concern is that it makes the test case to be potentially brittle probably:

If you have --implicit-check-not="Generic Options:", then if somebody deside to change "Option"->"option",
then you'll get a broken test case. Not sure if we should think about this scenario in that way though.

grimar: Note: I am not sure why "Options" is in upper case when "Generic", but lower case when…
jhendersonAuthorUnsubmitted
Done
ReplyInline Actions

I agree that the inconsistent case is bad. I'm not sure of a good alternative (apart from making it a regex pattern of [Oo]). I'll give it a bit more thought.

jhenderson: I agree that the inconsistent case is bad. I'm not sure of a good alternative (apart from…
jhendersonAuthorUnsubmitted
Done
ReplyInline Actions

I've tried making it less likely to get missed, by putting the -NOT checks out-of-line, so that they're next to the positive checks. If the case is changed, hopefully the person updating the test will change both positive and negative cases.

jhenderson: I've tried making it less likely to get missed, by putting the -NOT checks out-of-line, so that…
CHECK: @FILE CHECK: @FILE

test/tools/llvm-strings/length.test

RUN: not llvm-strings -n 0 2>&1 %S/Inputs/variable-length | FileCheck --check-prefix CHECK-0 %s # Show that llvm-strings prints only strings with length of at least the
RUN: llvm-strings -n 1 %S/Inputs/variable-length | FileCheck --check-prefix CHECK-1 %s # requested number of bytes.
RUN: llvm-strings -n 2 %S/Inputs/variable-length | FileCheck --check-prefix CHECK-2 %s
RUN: llvm-strings %S/Inputs/variable-length | FileCheck --check-prefix CHECK-DEFAULT %s
STDIN: a
STDIN: ab
STDIN: abc
STDIN: abcd
CHECK-DEFAULT: abcd RUN: echo a > %t
RUN: echo ab >> %t
RUN: echo abc >> %t
RUN: echo abcd >> %t
RUN: echo abcde >> %t
RUN: not llvm-strings -n 0 2>&1 %t | FileCheck --check-prefix CHECK-0 %s
RUN: llvm-strings -n 1 %t | FileCheck --check-prefix CHECK-1 %s --implicit-check-not={{.}}
RUN: llvm-strings -n 2 %t | FileCheck --check-prefix CHECK-2 %s --implicit-check-not={{.}}
RUN: llvm-strings -n 4 %t | FileCheck --check-prefix CHECK-4 %s --implicit-check-not={{.}}
RUN: llvm-strings -n 5 %t | FileCheck --check-prefix CHECK-5 %s --implicit-check-not={{.}}
grimarUnsubmitted
Done
ReplyInline Actions

What if -n 6 or -n -1 or -n foo?

grimar: What if `-n 6` or `-n -1` or `-n foo`?
jhendersonAuthorUnsubmitted
Done
ReplyInline Actions

-n -1 is actually broken. It looks like the command-line parser converts it to an unsigned, resulting in it being the equivalent of UINT_MAX. As GNU strings allows -1 as an alias for -n 1, we probably don't want to try to address this immediately, and fix that issue first, then see if the problem is still possible afterwards.

I'll add -n foo and -n 6 cases.

jhenderson: `-n -1` is actually broken. It looks like the command-line parser converts it to an unsigned…
jhendersonAuthorUnsubmitted
Done
ReplyInline Actions

-n -1 is actually broken. It looks like the command-line parser converts it to an unsigned

Turns out it's an int in the parser, so the parser accepts it, but it is then static_cast to a size_t! Anyway, still broken, and I've filed a bug for negative arguments.

jhenderson: > -n -1 is actually broken. It looks like the command-line parser converts it to an unsigned…
# Default is equivalent to -n 4.
RUN: llvm-strings %t | FileCheck --check-prefix CHECK-4 %s --implicit-check-not={{.}}
# Show --bytes works too.
RUN: llvm-strings --bytes 2 %t | FileCheck --check-prefix CHECK-2 %s --implicit-check-not={{.}}
# Show different syntaxes work.
RUN: llvm-strings --bytes=2 %t | FileCheck --check-prefix CHECK-2 %s --implicit-check-not={{.}}
RUN: llvm-strings -n=2 %t | FileCheck --check-prefix CHECK-2 %s --implicit-check-not={{.}}
CHECK-0: invalid minimum string length 0 CHECK-0: invalid minimum string length 0
CHECK-1: a CHECK-1: a
CHECK-1: ab CHECK-1-NEXT: ab
CHECK-1: abc CHECK-1-NEXT: abc
CHECK-1: abcd CHECK-1-NEXT: abcd
CHECK-1-NEXT: abcde
CHECK-2-NOT: a
CHECK-2: ab CHECK-2: ab
CHECK-2: abc CHECK-2-NEXT: abc
CHECK-2: abcd CHECK-2-NEXT: abcd
CHECK-2-NEXT: abcde
CHECK-4: abcd
CHECK-4-NEXT: abcde
CHECK-5: abcde

test/tools/llvm-strings/multiple-inputs.test

# Show that llvm-strings reads every file specified on the command-line.
# This test uses --print-file-name and --radix to show that the values are
# updated for the second file.
RUN: echo abcd > %t1
RUN: echo 1234 >> %t1
RUN: echo defg > %t2
RUN: llvm-strings --print-file-name --radix=d %t1 %t2 \
RUN: | FileCheck %s -DFILE1=%t1 -DFILE2=%t2 --implicit-check-not={{.}}
CHECK: [[FILE1]]: 0 abcd
CHECK-NEXT: [[FILE1]]: 5 1234
CHECK-NEXT: [[FILE2]]: 0 defg

test/tools/llvm-strings/negative-char.test

## Show that llvm-strings can handle a negative signed char value (i.e. > 0x7f).
## Such characters should form string delimiters like other unprintable ones.
# RUN: echo -e "z\0\x80\0a\0" | llvm-strings --bytes 1 - | FileCheck %s # RUN: echo -e "z\0\x80\0a\0" | llvm-strings --bytes 1 - | FileCheck %s
# CHECK: z{{$}} # CHECK: z{{$}}
# CHECK-NEXT: {{^}}a # CHECK-NEXT: {{^}}a

test/tools/llvm-strings/nested-archives.test

RUN: rm -rf %t && mkdir -p %t
RUN: llvm-ar --format=gnu crs %t/inner.a %S/Inputs/abcd
RUN: llvm-ar --format=gnu crs %t/outer.a %t/inner.a
RUN: llvm-strings %t/outer.a | FileCheck %s
CHECK: !<arch>
CHECK: inner.a/ 0 0 0 644 72 `
CHECK: !<arch>
CHECK: abcd/ 0 0 0 644 4 `
CHECK: abcd

test/tools/llvm-strings/option-grouping.test

# Show that llvm-strings can handle grouped short options.
RUN: echo abcd > %t
RUN: llvm-strings -af %t | FileCheck %s -DFILE=%t
CHECK: [[FILE]]: abcd

test/tools/llvm-strings/radix-filename.test

RUN: llvm-strings --print-file-name %S/Inputs/numbers \ # Show that --radix combined with --print-file-name produces correct output.
RUN: | FileCheck %s --check-prefix CHECK-NONE
RUN: llvm-strings --print-file-name -t d %S/Inputs/numbers \ RUN: echo one > %t
RUN: | FileCheck %s --check-prefix CHECK-DEC --strict-whitespace RUN: echo two >> %t
RUN: llvm-strings --print-file-name -t o %S/Inputs/numbers \ RUN: echo three >> %t
RUN: | FileCheck %s --check-prefix CHECK-OCT --strict-whitespace RUN: echo four >> %t
RUN: llvm-strings --print-file-name -t x %S/Inputs/numbers \ RUN: echo five >> %t
RUN: | FileCheck %s --check-prefix CHECK-HEX --strict-whitespace RUN: echo six >> %t
RUN: echo seven >> %t
CHECK-NONE: numbers: three RUN: echo eight >> %t
CHECK-NONE: numbers: four RUN: echo nine >> %t
CHECK-NONE: numbers: five RUN: echo ten >> %t
CHECK-NONE: numbers: seven
CHECK-NONE: numbers: eight RUN: llvm-strings --print-file-name %t \
CHECK-NONE: numbers: nine RUN: | FileCheck %s --check-prefix CHECK-NONE -DFILE=%t --implicit-check-not={{.}}
RUN: llvm-strings --print-file-name -t d %t \
CHECK-DEC: numbers: 8 three RUN: | FileCheck %s --check-prefix CHECK-DEC --strict-whitespace -DFILE=%t --implicit-check-not={{.}}
CHECK-DEC: numbers: 14 four RUN: llvm-strings --print-file-name -t o %t \
CHECK-DEC: numbers: 19 five RUN: | FileCheck %s --check-prefix CHECK-OCT --strict-whitespace -DFILE=%t --implicit-check-not={{.}}
CHECK-DEC: numbers: 28 seven RUN: llvm-strings --print-file-name -t x %t \
CHECK-DEC: numbers: 34 eight RUN: | FileCheck %s --check-prefix CHECK-HEX --strict-whitespace -DFILE=%t --implicit-check-not={{.}}
CHECK-DEC: numbers: 40 nine
CHECK-NONE: [[FILE]]: three
CHECK-OCT: numbers: 10 three CHECK-NONE: [[FILE]]: four
CHECK-OCT: numbers: 16 four CHECK-NONE: [[FILE]]: five
CHECK-OCT: numbers: 23 five CHECK-NONE: [[FILE]]: seven
CHECK-OCT: numbers: 34 seven CHECK-NONE: [[FILE]]: eight
CHECK-OCT: numbers: 42 eight CHECK-NONE: [[FILE]]: nine
CHECK-OCT: numbers: 50 nine
CHECK-DEC: [[FILE]]: 8 three
CHECK-HEX: numbers: 8 three CHECK-DEC: [[FILE]]: 14 four
CHECK-HEX: numbers: e four CHECK-DEC: [[FILE]]: 19 five
CHECK-HEX: numbers: 13 five CHECK-DEC: [[FILE]]: 28 seven
CHECK-HEX: numbers: 1c seven CHECK-DEC: [[FILE]]: 34 eight
CHECK-HEX: numbers: 22 eight CHECK-DEC: [[FILE]]: 40 nine
CHECK-HEX: numbers: 28 nine
CHECK-OCT: [[FILE]]: 10 three
CHECK-OCT: [[FILE]]: 16 four
CHECK-OCT: [[FILE]]: 23 five
CHECK-OCT: [[FILE]]: 34 seven
CHECK-OCT: [[FILE]]: 42 eight
CHECK-OCT: [[FILE]]: 50 nine
CHECK-HEX: [[FILE]]: 8 three
CHECK-HEX: [[FILE]]: e four
CHECK-HEX: [[FILE]]: 13 five
CHECK-HEX: [[FILE]]: 1c seven
CHECK-HEX: [[FILE]]: 22 eight
CHECK-HEX: [[FILE]]: 28 nine

test/tools/llvm-strings/radix.test

RUN: llvm-strings %S/Inputs/numbers | FileCheck %s -check-prefix CHECK-NONE # Show that llvm-strings can handle the -t/--radix switch properly.
RUN: llvm-strings -t d %S/Inputs/numbers | FileCheck %s -check-prefix CHECK-DEC --strict-whitespace
RUN: llvm-strings -t o %S/Inputs/numbers | FileCheck %s -check-prefix CHECK-OCT --strict-whitespace RUN: echo one > %t
RUN: llvm-strings -t x %S/Inputs/numbers | FileCheck %s -check-prefix CHECK-HEX --strict-whitespace RUN: echo two >> %t
RUN: echo three >> %t
RUN: echo four >> %t
RUN: echo five >> %t
RUN: echo six >> %t
RUN: echo seven >> %t
RUN: echo eight >> %t
RUN: echo nine >> %t
RUN: echo ten >> %t
RUN: llvm-strings %t | FileCheck %s -check-prefix CHECK-NONE --implicit-check-not={{.}}
RUN: llvm-strings -t d %t | FileCheck %s -check-prefix CHECK-DEC --strict-whitespace --implicit-check-not={{.}}
RUN: llvm-strings -t o %t | FileCheck %s -check-prefix CHECK-OCT --strict-whitespace --implicit-check-not={{.}}
RUN: llvm-strings -t x %t | FileCheck %s -check-prefix CHECK-HEX --strict-whitespace --implicit-check-not={{.}}
# Show --radix works too.
RUN: llvm-strings --radix d %t | FileCheck %s -check-prefix CHECK-DEC --strict-whitespace
RUN: llvm-strings --radix o %t | FileCheck %s -check-prefix CHECK-OCT --strict-whitespace
RUN: llvm-strings --radix x %t | FileCheck %s -check-prefix CHECK-HEX --strict-whitespace
# Show different syntaxes work.
RUN: llvm-strings --radix=d %t | FileCheck %s -check-prefix CHECK-DEC --strict-whitespace
RUN: llvm-strings -t=d %t | FileCheck %s -check-prefix CHECK-DEC --strict-whitespace
CHECK-NONE: {{^}}three CHECK-NONE: {{^}}three
CHECK-NONE: {{^}}four CHECK-NONE: {{^}}four
CHECK-NONE: {{^}}five CHECK-NONE: {{^}}five
CHECK-NONE: {{^}}seven CHECK-NONE: {{^}}seven
CHECK-NONE: {{^}}eight CHECK-NONE: {{^}}eight
CHECK-NONE: {{^}}nine CHECK-NONE: {{^}}nine
Show All 12 Lines
CHECK-OCT: {{^}} 50 nine CHECK-OCT: {{^}} 50 nine
CHECK-HEX: {{^}} 8 three CHECK-HEX: {{^}} 8 three
CHECK-HEX: {{^}} e four CHECK-HEX: {{^}} e four
CHECK-HEX: {{^}} 13 five CHECK-HEX: {{^}} 13 five
CHECK-HEX: {{^}} 1c seven CHECK-HEX: {{^}} 1c seven
CHECK-HEX: {{^}} 22 eight CHECK-HEX: {{^}} 22 eight
CHECK-HEX: {{^}} 28 nine CHECK-HEX: {{^}} 28 nine
# Show that an invalid value is rejected.
RUN: not llvm-strings --radix z %t 2>&1 | FileCheck %s --check-prefix=INVALID
INVALID: llvm-strings{{.*}}: for the --radix option: Cannot find option named 'z'!

test/tools/llvm-strings/response-file.test

# Show that llvm-strings can consume a response file.
RUN: echo abcd > %t
RUN: echo --print-file-name %t > %t.rsp
RUN: llvm-strings --radix=d @%t.rsp | FileCheck %s -DFILE=%t
CHECK: [[FILE]]: 0 abcd

test/tools/llvm-strings/stdin-filename.test

# Show that --print-file-name works when the input is standard input.
RUN: echo abcd | llvm-strings -f - | FileCheck %s RUN: echo abcd | llvm-strings -f - | FileCheck %s
RUN: echo abcd | llvm-strings -f | FileCheck %s
RUN: echo abcd | llvm-strings --print-file-name - | FileCheck %s RUN: echo abcd | llvm-strings --print-file-name - | FileCheck %s
RUN: echo abcd | llvm-strings --print-file-name | FileCheck %s
CHECK: {standard input}: abcd CHECK: {standard input}: abcd

test/tools/llvm-strings/stdin.test

# Show that llvm-strings can handle stdin input properly.
# Case 1: output with single string.
RUN: echo -n abcdefg | llvm-strings - | FileCheck %s --check-prefix=CASE1 --implicit-check-not={{.}}
CASE1: abcdefg
# Case 2: output too short for string.
RUN: echo -n abc | llvm-strings - | FileCheck %s --implicit-check-not={{.}} --allow-empty
# Case 3: output with new line.
RUN: echo -e abcd\nefgh | llvm-strings - | FileCheck %s --check-prefix=CASE3 --implicit-check-not={{.}}
CASE3: abcd
jhendersonAuthorUnsubmitted
Done
ReplyInline Actions

On my Linux box, "echo -e" doesn't work without quoting the string to be echoed, so I did that throughout this test.

jhenderson: On my Linux box, "echo -e" doesn't work without quoting the string to be echoed, so I did that…
CASE3-NEXT: efgh
# Case 4: output containing unprintable characters.
RUN: echo -e abcd\x00ef\x1fghij\x7fklmn | llvm-strings - | FileCheck %s --check-prefix=CASE4 --implicit-check-not={{.}}
CASE4: abcd
CASE4-NEXT: ghij
CASE4-NEXT: klmn
# Case 5: no file name specified is equivalent to explicitly requesting stdin.
RUN: echo -n abcdefg | llvm-strings | FileCheck %s --check-prefix=CASE1 --implicit-check-not={{.}}

test/tools/llvm-strings/terminator-neg.test

RUN: echo -n abc | llvm-strings - | FileCheck -allow-empty %s
CHECK-NOT: abc

test/tools/llvm-strings/terminator.test

RUN: echo -n abcdefg | llvm-strings - | FileCheck %s
CHECK: abcdefg

test/tools/llvm-strings/version.test

# Show that --version works for llvm-strings.
grimarUnsubmitted
Not Done
ReplyInline Actions

I.d probably just put this into help.test. What do you think?

grimar: I.d probably just put this into `help.test`. What do you think?
jhendersonAuthorUnsubmitted
Done
ReplyInline Actions

I prefer to keep each switch to a separate test, unless they're intricately linked (in this case --help and --version are not really linked).

jhenderson: I prefer to keep each switch to a separate test, unless they're intricately linked (in this…
RUN: llvm-strings --version | FileCheck %s
CHECK: LLVM version
rupprechtUnsubmitted
Done
ReplyInline Actions

This is actually testing the default cmake config; see this comment from llvm/test/tools/llvm-readobj/basic.test:

# Test version switch.
RUN: llvm-readobj --version | FileCheck %s --check-prefix=VERSION
RUN: llvm-readelf --version | FileCheck %s --check-prefix=VERSION
# In default configuration we could match "LLVM version", but the "LLVM" part
# can be changed with PACKAGE_NAME in CMake, so we match only version.
VERSION: version
rupprecht: This is actually testing the default cmake config; see this comment from `llvm/test/tools/llvm…
jhendersonAuthorUnsubmitted
Done
ReplyInline Actions

Thanks! I'll fix that.

jhenderson: Thanks! I'll fix that.

test/tools/llvm-strings/whitespace.test

RUN: echo -n abcd | llvm-strings - | FileCheck %s --strict-whitespace # Show that the default output format matches GNU strings.
RUN: echo -n abcd | llvm-strings - | FileCheck %s --strict-whitespace --implicit-check-not={{.}}
CHECK: {{^}}abcd{{$}} CHECK: {{^}}abcd{{$}}
test/tools/llvm-strings/stdin.test